16_music_classification_light_ryazhskikh.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "16_music_classification_light_ryazhskikh.ipynb",
      "version": "0.3.2",
      "provenance": [],
      "collapsed_sections": [
        "ChRHH3IF-caf"
      ],
      "toc_visible": true,
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/darja/NeuralUni/blob/master/16_music_classification_light_ryazhskikh.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "QxKQMFQ5mTAF",
        "colab_type": "text"
      },
      "source": [
        "# Классификация музыки по жанрам"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "hP3axDhyjqMv",
        "colab_type": "text"
      },
      "source": [
        "**Задание light:**\n",
        "\n",
        "Для базы GTZAN Genre Collection\n",
        "1. Сделать классификацию с помощью полносвязной нейронной сети. Найти оптимальную архитектуру (количество слоев и нейронов) и оптимальный набор гипермапараметров (оптимизатор, размер мини-выборки, количество эпох и т.д.).\n",
        "2. Используя библиотеку librosa (http://librosa.github.io/librosa/feature.html) поварьировать набор признаков, извлекаемых из аудио-файла. В качестве примера некоторые параметры можно вычислять не от всего файла длительностью 30 секунд, а от его частей, например длительностью 10 секунд. Также можно обратить внимание на количество mfcc коэффициентов.\n",
        "3. (Опционально) Применить линейные модели (решающие деревья, градиентный бустинг) из библиотеки sklearn на полученном датасете, сделать ансамбль моделей из нейросети и линейных моделей.\n",
        " \n",
        "**Задание pro:**\n",
        "1. Выполните задание light.\n",
        "2. Получите визуальные представления аудио файлов (возможно длительности менее 30 секунд).\n",
        "3. Постройте модель классификации с использованием сверточных сетей. \n",
        "Удалось ли превзойти полносвязные сети?"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "RF1ra7vQmcYp",
        "colab_type": "text"
      },
      "source": [
        "# Подготовка окружения"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6UfGFM-6mbnb",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# feature extractoring and preprocessing data\n",
        "import librosa\n",
        "import cv2\n",
        "import pandas as pd\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "#matplotlib inline\n",
        "import os\n",
        "import PIL\n",
        "from IPython.display import display\n",
        "from PIL import Image\n",
        "import pathlib\n",
        "import csv\n",
        "\n",
        "# Preprocessing\n",
        "from sklearn.model_selection import train_test_split\n",
        "from sklearn.preprocessing import LabelEncoder, StandardScaler\n",
        "from tensorflow.keras.preprocessing import image\n",
        "\n",
        "# Google\n",
        "from google.colab import files\n",
        "from google.colab import drive\n",
        "\n",
        "\n",
        "#Keras\n",
        "from tensorflow.keras.models import Sequential\n",
        "from tensorflow.keras.layers import Dense, Dropout, BatchNormalization\n",
        "from tensorflow.keras import callbacks\n",
        "from tensorflow.keras.optimizers import Adam\n",
        "from tensorflow.keras.wrappers.scikit_learn import KerasClassifier\n",
        "\n",
        "import warnings\n",
        "warnings.filterwarnings('ignore')\n",
        "\n",
        "genres = ['blues', 'classical', 'country', 'disco', 'hiphop', 'jazz', 'metal', 'pop', 'reggae', 'rock']\n",
        "\n",
        "slice_duration = 5\n",
        "features_file = f\"sound_features_{slice_duration}sec.csv\""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "IRq9qtMnmhdA",
        "colab_type": "text"
      },
      "source": [
        "## Подключение Google Drive"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RDvFcj3CjnxG",
        "colab_type": "code",
        "outputId": "ab68a84f-bd6d-4f58-bf81-ece8797fb6b3",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 122
        }
      },
      "source": [
        "drive.mount('/content/drive')\n",
        "#http://marsyas.info/downloads/datasets.html"
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Go to this URL in a browser: https://accounts.google.com/o/oauth2/auth?client_id=947318989803-6bn6qk8qdgf4n4g3pfee6491hc0brc4i.apps.googleusercontent.com&redirect_uri=urn%3Aietf%3Awg%3Aoauth%3A2.0%3Aoob&scope=email%20https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fdocs.test%20https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fdrive%20https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fdrive.photos.readonly%20https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fpeopleapi.readonly&response_type=code\n",
            "\n",
            "Enter your authorization code:\n",
            "··········\n",
            "Mounted at /content/drive\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "JiaNQ_TNm0mG",
        "colab_type": "text"
      },
      "source": [
        "# Загрузка наборов данных"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "99LRhO3XObUK",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "!cp /content/drive/'My Drive'/neural/genres.zip genres.zip\n",
        "!unzip genres.zip"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "PjBe-Phn0TRD",
        "colab_type": "text"
      },
      "source": [
        "## Утилиты"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Pbp5l7_ed7d_",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def append_features_header(header, feature_name):\n",
        "    if (feature_name == \"mfcc\"):\n",
        "        header = np.append(header, np.vectorize(lambda i: f'mfcc{i}')(np.arange(1, 21)))\n",
        "    elif (feature_name == \"tonnetz\"):\n",
        "        header = np.append(header, np.vectorize(lambda i: f'tonnetz{i}')(np.arange(1, 7)))\n",
        "    elif (feature_name == \"chroma_cqt\"):\n",
        "        header = np.append(header, np.vectorize(lambda i: f'chroma_cqt{i}')(np.arange(1, 13)))\n",
        "    else:\n",
        "        header = np.append(header, feature_name)\n",
        "    return header\n",
        "    \n",
        "def create_features_header(features):\n",
        "    header = [\"filename\"]\n",
        "    for feature_name in features:\n",
        "        header = append_features_header(header, feature_name)\n",
        "    header = np.append(header, 'label')\n",
        "    return header\n",
        "\n",
        "def get_feature_by_name(feature_name, y, sr):\n",
        "    if feature_name == \"chroma_stft\":\n",
        "        return librosa.feature.chroma_stft(y=y, sr=sr)\n",
        "    elif feature_name == \"rmse\":\n",
        "        return librosa.feature.rmse(y=y)\n",
        "    elif feature_name == \"spectral_centroid\":\n",
        "        return librosa.feature.spectral_centroid(y=y, sr=sr)\n",
        "    elif feature_name == \"spectral_bandwidth\":\n",
        "        return librosa.feature.spectral_bandwidth(y=y, sr=sr)\n",
        "    elif feature_name == \"rolloff\":\n",
        "        return librosa.feature.spectral_rolloff(y=y, sr=sr)\n",
        "    elif feature_name == \"zero_crossing_rate\":\n",
        "        return librosa.feature.zero_crossing_rate(y)\n",
        "    elif feature_name == \"mfcc\":\n",
        "        return librosa.feature.mfcc(y=y, sr=sr)\n",
        "    elif feature_name == 'spectral_flatness':\n",
        "        return librosa.feature.spectral_flatness(y = y)\n",
        "    elif feature_name == 'tonnetz':\n",
        "        return librosa.feature.tonnetz(y=y, sr=sr)\n",
        "    elif feature_name == 'tempogram':\n",
        "        return librosa.feature.tempogram(y=y, sr=sr)\n",
        "    elif feature_name == 'chroma_cqt':\n",
        "        return librosa.feature.chroma_cqt(y=y, sr=sr)\n",
        "    elif feature_name == 'melspectrogram':\n",
        "        return librosa.feature.melspectrogram(y=y, sr=sr)\n",
        "    else:\n",
        "        print(\"Incorrect feature: \", feature_name)\n",
        "        \n",
        "def prepare_feature_value(feature_name, feature_value):\n",
        "#     print(f'{feature_name} value shape: {feature_value.shape})')\n",
        "    if (feature_name in ['mfcc', 'tonnetz', 'chroma_cqt']):\n",
        "        return np.apply_along_axis(lambda x: np.mean(x), 1, feature_value)\n",
        "    else:\n",
        "        return np.mean(feature_value)\n",
        "        \n",
        "def get_features(y, sr, features):\n",
        "    features_values = []\n",
        "    for feature_name in features:\n",
        "        feature = get_feature_by_name(feature_name, y, sr)\n",
        "        value = prepare_feature_value(feature_name, feature)\n",
        "        features_values = np.append(features_values, value)\n",
        "    return features_values\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "SLI5UMKspxGc",
        "colab_type": "text"
      },
      "source": [
        "## Формирование датасета с характеристиками звука"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "GcjMDOz57pFY",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "0746a7c1-64db-4fe6-af1e-98d3571f474c"
      },
      "source": [
        "print(header)"
      ],
      "execution_count": 44,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "['filename', 'chroma_stft', 'rmse', 'spectral_centroid', 'spectral_bandwidth', 'rolloff', 'zero_crossing_rate', 'label']\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Jwb8rGCY1-G2",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 204
        },
        "outputId": "81bc2ed6-59e6-47bb-d008-67474a37bd27"
      },
      "source": [
        "features = ['chroma_stft', \n",
        "            'rmse',\n",
        "            'tonnetz',\n",
        "            'spectral_centroid', \n",
        "            'spectral_bandwidth', \n",
        "            'spectral_flatness', \n",
        "            'rolloff', \n",
        "            'zero_crossing_rate', \n",
        "            'chroma_cqt',\n",
        "            'melspectrogram',\n",
        "            'tempogram',\n",
        "            'mfcc'\n",
        "]\n",
        "header = create_features_header(features)\n",
        "file = open(features_file, 'w', newline='')\n",
        "with file:\n",
        "    writer = csv.writer(file)\n",
        "    writer.writerow(header)\n",
        "    \n",
        "parts = np.arange(0, 30 // slice_duration)\n",
        "\n",
        "file = open(features_file, 'a', newline='')\n",
        "with file:\n",
        "    writer = csv.writer(file)\n",
        "\n",
        "    for genre in genres:\n",
        "        dir_content = os.listdir(f'./genres/{genre}')\n",
        "        i = 1\n",
        "        n = len(dir_content)\n",
        "        print()\n",
        "        for filename in dir_content:\n",
        "            songname = f'./genres/{genre}/{filename}'\n",
        "            print(\"\\r\", end = f'{genre}: {i} of {n}')\n",
        "            i += 1\n",
        "            \n",
        "            for part in parts:\n",
        "                y, sr = librosa.load(songname, mono=True, offset = part * slice_duration, duration=slice_duration)\n",
        "\n",
        "                row = [f\"{filename}, {part * slice_duration}\"]\n",
        "                row = np.append(row, get_features(y, sr, features))\n",
        "                row = np.append(row, genre)\n",
        "                writer.writerow(row)\n",
        "    print()"
      ],
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "\n",
            "blues: 100 of 100\n",
            "classical: 100 of 100\n",
            "country: 100 of 100\n",
            "disco: 100 of 100\n",
            "hiphop: 100 of 100\n",
            "jazz: 100 of 100\n",
            "metal: 100 of 100\n",
            "pop: 100 of 100\n",
            "reggae: 100 of 100\n",
            "rock: 100 of 100\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "wDD-eMDHqshh",
        "colab_type": "text"
      },
      "source": [
        "## Загрузка датасета с характеристиками звука"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5FSrTvmqtWMx",
        "colab_type": "code",
        "outputId": "be62de4a-b0d6-4776-a215-635f91d70b39",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 68
        }
      },
      "source": [
        "features = create_features_header(['chroma_stft', \n",
        "            'rmse',\n",
        "            'tonnetz',\n",
        "            'spectral_centroid', \n",
        "            'spectral_bandwidth', \n",
        "            'spectral_flatness', \n",
        "            'rolloff', \n",
        "            'zero_crossing_rate', \n",
        "            'chroma_cqt',\n",
        "            'melspectrogram',\n",
        "            'tempogram',\n",
        "            'mfcc'\n",
        "])\n",
        "\n",
        "data = pd.read_csv(features_file)\n",
        "data = data[features]\n",
        "\n",
        "print(data.shape)\n",
        "# Dropping unneccesary columns\n",
        "data = data.drop(['filename'],axis=1)\n",
        "\n",
        "#Encoding the Labels\n",
        "genre_list = data.iloc[:, -1]\n",
        "encoder = LabelEncoder()\n",
        "y = encoder.fit_transform(genre_list)\n",
        "data = data.drop(['label'],axis=1)\n",
        "\n",
        "#Scaling the Feature columns\n",
        "scaler = StandardScaler()\n",
        "\n",
        "X = scaler.fit_transform(np.array(data.iloc[:, :], dtype = float))\n",
        "\n",
        "#Dividing data into training and Testing set\n",
        "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.05)\n",
        "\n",
        "print(\"Train set: \", X_train.shape)\n",
        "print(\"Test set: \", X_test.shape)"
      ],
      "execution_count": 38,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(6000, 49)\n",
            "Train set:  (5700, 47)\n",
            "Test set:  (300, 47)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Zk5nKnXu7ase",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "491443be-344a-4098-b3ce-cf50d6453a9d"
      },
      "source": [
        "scaler"
      ],
      "execution_count": 39,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "StandardScaler(copy=True, with_mean=True, with_std=True)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 39
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "vAQ_C8kVTr9T",
        "colab_type": "text"
      },
      "source": [
        "# Подготовка классификаторов"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "2rYS6nVKTk9A",
        "colab_type": "text"
      },
      "source": [
        "## Классификация методами машинного обучения"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fyXonwHHORiW",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from sklearn import metrics\n",
        "\n",
        "from sklearn.multioutput import MultiOutputClassifier\n",
        "from sklearn.ensemble import RandomForestClassifier, AdaBoostClassifier\n",
        "from sklearn.tree import DecisionTreeClassifier\n",
        "\n",
        "from keras.utils import to_categorical\n",
        "\n",
        "y_train_ohe = to_categorical(y_train)\n",
        "y_test_ohe = to_categorical(y_test)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yG_VGIVwUOmO",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "d004d3ac-e8db-45c2-d9a1-d6045a788fcb"
      },
      "source": [
        "# bdt = AdaBoostClassifier(DecisionTreeClassifier(max_depth=3), n_estimators=100)\n",
        "\n",
        "forest = RandomForestClassifier(n_estimators=100, random_state=3)\n",
        "bdt = AdaBoostClassifier(forest, n_estimators=50)\n",
        "\n",
        "multi_target_forest = MultiOutputClassifier(bdt, n_jobs=1)\n",
        "classifier = multi_target_forest.fit(X_train, y_train_ohe)\n",
        "\n",
        "test_predictions = classifier.predict(X_test)\n",
        "acc = metrics.accuracy_score(y_test_ohe, test_predictions)\n",
        "      \n",
        "print(\"Accuracy: \", acc)"
      ],
      "execution_count": 146,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Accuracy:  0.5466666666666666\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "H9UrExqYppT-",
        "colab_type": "text"
      },
      "source": [
        "## Классификация с помощью нейросети"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ydxD84vYBZ0p",
        "colab_type": "code",
        "outputId": "b0ec6dcc-89db-4760-8751-b56e2db4303f",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1258
        }
      },
      "source": [
        "# Коллбэки\n",
        "model_callbacks = [\n",
        "    callbacks.ReduceLROnPlateau(monitor='val_loss', factor=0.7, patience=3, min_lr=1e-6, verbose=1)\n",
        "]\n",
        "\n",
        "optimizer = Adam(lr=0.01)\n",
        "\n",
        "model = Sequential()\n",
        "model.add(Dense(256, activation='relu', input_shape=(X_train.shape[1],), kernel_initializer=\"glorot_uniform\"))\n",
        "model.add(Dropout(0.1))\n",
        "model.add(Dense(10, activation='softmax'))\n",
        "\n",
        "model.compile(optimizer=optimizer,\n",
        "              loss='sparse_categorical_crossentropy',\n",
        "              metrics=['accuracy'])\n",
        "\n",
        "history = model.fit(X_train, y_train,\n",
        "                    epochs = 30,\n",
        "                    batch_size = 16,\n",
        "                    validation_split = 0.1,\n",
        "                    callbacks = model_callbacks)"
      ],
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 5130 samples, validate on 570 samples\n",
            "Epoch 1/30\n",
            "5130/5130 [==============================] - 1s 202us/sample - loss: 1.2726 - acc: 0.5811 - val_loss: 0.9586 - val_acc: 0.6702\n",
            "Epoch 2/30\n",
            "5130/5130 [==============================] - 1s 172us/sample - loss: 0.9715 - acc: 0.6949 - val_loss: 0.9272 - val_acc: 0.7035\n",
            "Epoch 3/30\n",
            "5130/5130 [==============================] - 1s 171us/sample - loss: 0.8558 - acc: 0.7292 - val_loss: 0.8743 - val_acc: 0.7105\n",
            "Epoch 4/30\n",
            "5130/5130 [==============================] - 1s 172us/sample - loss: 0.7653 - acc: 0.7686 - val_loss: 0.8577 - val_acc: 0.7561\n",
            "Epoch 5/30\n",
            "5130/5130 [==============================] - 1s 173us/sample - loss: 0.6577 - acc: 0.7953 - val_loss: 0.7461 - val_acc: 0.7684\n",
            "Epoch 6/30\n",
            "5130/5130 [==============================] - 1s 171us/sample - loss: 0.6272 - acc: 0.8129 - val_loss: 0.6605 - val_acc: 0.8193\n",
            "Epoch 7/30\n",
            "5130/5130 [==============================] - 1s 169us/sample - loss: 0.5620 - acc: 0.8363 - val_loss: 0.8799 - val_acc: 0.7632\n",
            "Epoch 8/30\n",
            "5130/5130 [==============================] - 1s 176us/sample - loss: 0.5989 - acc: 0.8271 - val_loss: 0.7082 - val_acc: 0.8140\n",
            "Epoch 9/30\n",
            "5088/5130 [============================>.] - ETA: 0s - loss: 0.5515 - acc: 0.8481\n",
            "Epoch 00009: ReduceLROnPlateau reducing learning rate to 0.006999999843537807.\n",
            "5130/5130 [==============================] - 1s 181us/sample - loss: 0.5495 - acc: 0.8485 - val_loss: 0.8812 - val_acc: 0.7947\n",
            "Epoch 10/30\n",
            "5130/5130 [==============================] - 1s 174us/sample - loss: 0.3172 - acc: 0.8984 - val_loss: 0.5356 - val_acc: 0.8667\n",
            "Epoch 11/30\n",
            "5130/5130 [==============================] - 1s 171us/sample - loss: 0.2308 - acc: 0.9238 - val_loss: 0.5976 - val_acc: 0.8421\n",
            "Epoch 12/30\n",
            "5130/5130 [==============================] - 1s 173us/sample - loss: 0.2016 - acc: 0.9327 - val_loss: 0.5400 - val_acc: 0.8544\n",
            "Epoch 13/30\n",
            "5056/5130 [============================>.] - ETA: 0s - loss: 0.2080 - acc: 0.9300\n",
            "Epoch 00013: ReduceLROnPlateau reducing learning rate to 0.004899999825283885.\n",
            "5130/5130 [==============================] - 1s 172us/sample - loss: 0.2076 - acc: 0.9300 - val_loss: 0.5674 - val_acc: 0.8649\n",
            "Epoch 14/30\n",
            "5130/5130 [==============================] - 1s 172us/sample - loss: 0.1622 - acc: 0.9485 - val_loss: 0.5034 - val_acc: 0.8772\n",
            "Epoch 15/30\n",
            "5130/5130 [==============================] - 1s 167us/sample - loss: 0.1131 - acc: 0.9649 - val_loss: 0.4655 - val_acc: 0.8895\n",
            "Epoch 16/30\n",
            "5130/5130 [==============================] - 1s 172us/sample - loss: 0.0931 - acc: 0.9727 - val_loss: 0.4383 - val_acc: 0.8982\n",
            "Epoch 17/30\n",
            "5130/5130 [==============================] - 1s 172us/sample - loss: 0.1283 - acc: 0.9600 - val_loss: 0.5879 - val_acc: 0.8789\n",
            "Epoch 18/30\n",
            "5130/5130 [==============================] - 1s 171us/sample - loss: 0.1317 - acc: 0.9583 - val_loss: 0.5126 - val_acc: 0.8930\n",
            "Epoch 19/30\n",
            "4976/5130 [============================>.] - ETA: 0s - loss: 0.1355 - acc: 0.9574\n",
            "Epoch 00019: ReduceLROnPlateau reducing learning rate to 0.0034300000406801696.\n",
            "5130/5130 [==============================] - 1s 175us/sample - loss: 0.1332 - acc: 0.9581 - val_loss: 0.4971 - val_acc: 0.8877\n",
            "Epoch 20/30\n",
            "5130/5130 [==============================] - 1s 173us/sample - loss: 0.0887 - acc: 0.9700 - val_loss: 0.5222 - val_acc: 0.8965\n",
            "Epoch 21/30\n",
            "5130/5130 [==============================] - 1s 170us/sample - loss: 0.0596 - acc: 0.9803 - val_loss: 0.4557 - val_acc: 0.9053\n",
            "Epoch 22/30\n",
            "4848/5130 [===========================>..] - ETA: 0s - loss: 0.0470 - acc: 0.9847\n",
            "Epoch 00022: ReduceLROnPlateau reducing learning rate to 0.002401000028476119.\n",
            "5130/5130 [==============================] - 1s 168us/sample - loss: 0.0486 - acc: 0.9842 - val_loss: 0.4789 - val_acc: 0.9035\n",
            "Epoch 23/30\n",
            "5130/5130 [==============================] - 1s 171us/sample - loss: 0.0376 - acc: 0.9881 - val_loss: 0.4542 - val_acc: 0.9105\n",
            "Epoch 24/30\n",
            "5130/5130 [==============================] - 1s 174us/sample - loss: 0.0320 - acc: 0.9912 - val_loss: 0.4876 - val_acc: 0.9000\n",
            "Epoch 25/30\n",
            "5072/5130 [============================>.] - ETA: 0s - loss: 0.0410 - acc: 0.9864\n",
            "Epoch 00025: ReduceLROnPlateau reducing learning rate to 0.0016807000851258634.\n",
            "5130/5130 [==============================] - 1s 171us/sample - loss: 0.0410 - acc: 0.9864 - val_loss: 0.4565 - val_acc: 0.9070\n",
            "Epoch 26/30\n",
            "5130/5130 [==============================] - 1s 171us/sample - loss: 0.0444 - acc: 0.9865 - val_loss: 0.4736 - val_acc: 0.9053\n",
            "Epoch 27/30\n",
            "5130/5130 [==============================] - 1s 172us/sample - loss: 0.0285 - acc: 0.9922 - val_loss: 0.4720 - val_acc: 0.9018\n",
            "Epoch 28/30\n",
            "5120/5130 [============================>.] - ETA: 0s - loss: 0.0248 - acc: 0.9934\n",
            "Epoch 00028: ReduceLROnPlateau reducing learning rate to 0.0011764900758862494.\n",
            "5130/5130 [==============================] - 1s 170us/sample - loss: 0.0247 - acc: 0.9934 - val_loss: 0.4454 - val_acc: 0.9070\n",
            "Epoch 29/30\n",
            "5130/5130 [==============================] - 1s 170us/sample - loss: 0.0279 - acc: 0.9918 - val_loss: 0.4900 - val_acc: 0.8965\n",
            "Epoch 30/30\n",
            "5130/5130 [==============================] - 1s 172us/sample - loss: 0.0214 - acc: 0.9942 - val_loss: 0.4022 - val_acc: 0.9140\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "9_aA5J6qEQiF",
        "colab_type": "text"
      },
      "source": [
        "### График обучения"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "PRajIK5S521s",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 269
        },
        "outputId": "c5778848-416f-433b-bdc8-aede00ec49c0"
      },
      "source": [
        "plt.plot(history.history['acc'], label='Train')\n",
        "plt.plot(history.history['val_acc'], label='Validation')\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAD8CAYAAACMwORRAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "AfVpyp1mEZIE",
        "colab_type": "text"
      },
      "source": [
        "### Проверка на тестовой выборке"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "qsLON4KcCo6K",
        "colab_type": "code",
        "outputId": "133f489c-5bef-44ab-dcb8-c64b8ed040ad",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 51
        }
      },
      "source": [
        "result = model.evaluate(X_test, y_test)\n",
        "print(result)"
      ],
      "execution_count": 14,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "300/300 [==============================] - 0s 58us/sample - loss: 0.4990 - acc: 0.8933\n",
            "[0.49899046262105307, 0.8933333]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "wo81-KPqEcud",
        "colab_type": "text"
      },
      "source": [
        "### Визуальное представление нейросети"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EddTjE5mEPVE",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "29fae328-2dee-4325-c097-f9434a289366"
      },
      "source": [
        "from keras.utils import plot_model\n",
        "plot_model(model, to_file='genres_dense.png', show_layer_names=False, show_shapes=True)"
      ],
      "execution_count": 124,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Using TensorFlow backend.\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ChRHH3IF-caf",
        "colab_type": "text"
      },
      "source": [
        "### Подбор параметров (сомнительный)"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ShiwSf22-f7M",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 32368
        },
        "outputId": "48c20137-45db-41d7-8901-b4455a3f7d77"
      },
      "source": [
        "from sklearn.model_selection import GridSearchCV\n",
        "from sklearn.model_selection import PredefinedSplit\n",
        "\n",
        "test_fold = []\n",
        "print(len(y_train))\n",
        "for i in range(len(y_train) * 9 // 10):\n",
        "    test_fold.append(-1)\n",
        "print(len(test_fold))\n",
        "for i in range(len(y_train) // 10):\n",
        "    test_fold.append(0)\n",
        "\n",
        "print(len(test_fold))\n",
        "    \n",
        "ps = PredefinedSplit(test_fold)\n",
        "\n",
        "def create_model(optimizer = \"adam\", init = \"glorot_uniform\"):\n",
        "    model = Sequential()\n",
        "    model.add(Dense(256, activation='relu', input_shape=(X.shape[1],), kernel_initializer=init))\n",
        "    model.add(Dropout(0.3))\n",
        "#     model.add(Dense(128, activation='relu'))\n",
        "#     model.add(BatchNormalization())\n",
        "#     model.add(Dropout(0.2))\n",
        "#     model.add(Dense(32, activation='relu'))\n",
        "#     model.add(Dropout(0.2))\n",
        "    model.add(Dense(10, activation='softmax'))\n",
        "\n",
        "    model.compile(optimizer=optimizer,\n",
        "                  loss='sparse_categorical_crossentropy',\n",
        "                  metrics=['accuracy']\n",
        "                 )\n",
        "    return model\n",
        "\n",
        "estimator = KerasClassifier(build_fn=create_model)\n",
        "\n",
        "# grid search epochs, batch size and optimizer\n",
        "optimizers = ['rmsprop', 'adam']\n",
        "# optimizers = ['adam']\n",
        "epochs = [50]\n",
        "batches = [16, 32, 64]\n",
        "init = ['glorot_uniform', 'normal', 'uniform']\n",
        "# init = ['glorot_uniform']\n",
        "\n",
        "param_grid = dict(optimizer=optimizers, epochs=epochs, batch_size=batches, init=init)\n",
        "grid = GridSearchCV(estimator=estimator, param_grid=param_grid, scoring='balanced_accuracy', cv=ps)\n",
        "\n",
        "grid_result = grid.fit(X_train, y_train)"
      ],
      "execution_count": 16,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "5700\n",
            "5130\n",
            "5700\n",
            "Epoch 1/50\n",
            "5130/5130 [==============================] - 1s 222us/sample - loss: 1.6470 - acc: 0.4131\n",
            "Epoch 2/50\n",
            "5130/5130 [==============================] - 1s 210us/sample - loss: 1.3391 - acc: 0.5212\n",
            "Epoch 3/50\n",
            "5130/5130 [==============================] - 1s 218us/sample - loss: 1.2532 - acc: 0.5509\n",
            "Epoch 4/50\n",
            "5130/5130 [==============================] - 1s 220us/sample - loss: 1.2003 - acc: 0.5688\n",
            "Epoch 5/50\n",
            "5130/5130 [==============================] - 1s 218us/sample - loss: 1.1533 - acc: 0.5924\n",
            "Epoch 6/50\n",
            "5130/5130 [==============================] - 1s 220us/sample - loss: 1.1174 - acc: 0.5992\n",
            "Epoch 7/50\n",
            "5130/5130 [==============================] - 1s 228us/sample - loss: 1.0787 - acc: 0.6209\n",
            "Epoch 8/50\n",
            "5130/5130 [==============================] - 1s 229us/sample - loss: 1.0546 - acc: 0.6242\n",
            "Epoch 9/50\n",
            "5130/5130 [==============================] - 1s 224us/sample - loss: 1.0374 - acc: 0.6402\n",
            "Epoch 10/50\n",
            "5130/5130 [==============================] - 1s 221us/sample - loss: 1.0122 - acc: 0.6437\n",
            "Epoch 11/50\n",
            "5130/5130 [==============================] - 1s 220us/sample - loss: 0.9884 - acc: 0.6550\n",
            "Epoch 12/50\n",
            "5130/5130 [==============================] - 1s 221us/sample - loss: 0.9735 - acc: 0.6612\n",
            "Epoch 13/50\n",
            "5130/5130 [==============================] - 1s 224us/sample - loss: 0.9640 - acc: 0.6651\n",
            "Epoch 14/50\n",
            "5130/5130 [==============================] - 1s 221us/sample - loss: 0.9493 - acc: 0.6694\n",
            "Epoch 15/50\n",
            "5130/5130 [==============================] - 1s 227us/sample - loss: 0.9266 - acc: 0.6799\n",
            "Epoch 16/50\n",
            "5130/5130 [==============================] - 1s 220us/sample - loss: 0.9280 - acc: 0.6901\n",
            "Epoch 17/50\n",
            "5130/5130 [==============================] - 1s 218us/sample - loss: 0.9092 - acc: 0.6904\n",
            "Epoch 18/50\n",
            "5130/5130 [==============================] - 1s 221us/sample - loss: 0.8978 - acc: 0.6815\n",
            "Epoch 19/50\n",
            "5130/5130 [==============================] - 1s 220us/sample - loss: 0.8997 - acc: 0.6949\n",
            "Epoch 20/50\n",
            "5130/5130 [==============================] - 1s 222us/sample - loss: 0.8684 - acc: 0.7060\n",
            "Epoch 21/50\n",
            "5130/5130 [==============================] - 1s 217us/sample - loss: 0.8613 - acc: 0.6971\n",
            "Epoch 22/50\n",
            "5130/5130 [==============================] - 1s 222us/sample - loss: 0.8566 - acc: 0.6977\n",
            "Epoch 23/50\n",
            "5130/5130 [==============================] - 1s 223us/sample - loss: 0.8633 - acc: 0.6979\n",
            "Epoch 24/50\n",
            "5130/5130 [==============================] - 1s 217us/sample - loss: 0.8509 - acc: 0.7129\n",
            "Epoch 25/50\n",
            "5130/5130 [==============================] - 1s 218us/sample - loss: 0.8384 - acc: 0.7144\n",
            "Epoch 26/50\n",
            "5130/5130 [==============================] - 1s 217us/sample - loss: 0.8254 - acc: 0.7187\n",
            "Epoch 27/50\n",
            "5130/5130 [==============================] - 1s 215us/sample - loss: 0.8208 - acc: 0.7181\n",
            "Epoch 28/50\n",
            "5130/5130 [==============================] - 1s 219us/sample - loss: 0.8220 - acc: 0.7259\n",
            "Epoch 29/50\n",
            "5130/5130 [==============================] - 1s 223us/sample - loss: 0.8265 - acc: 0.7185\n",
            "Epoch 30/50\n",
            "5130/5130 [==============================] - 1s 218us/sample - loss: 0.8053 - acc: 0.7238\n",
            "Epoch 31/50\n",
            "5130/5130 [==============================] - 1s 218us/sample - loss: 0.7963 - acc: 0.7255\n",
            "Epoch 32/50\n",
            "5130/5130 [==============================] - 1s 226us/sample - loss: 0.7954 - acc: 0.7281\n",
            "Epoch 33/50\n",
            "5130/5130 [==============================] - 1s 220us/sample - loss: 0.7778 - acc: 0.7384\n",
            "Epoch 34/50\n",
            "5130/5130 [==============================] - 1s 227us/sample - loss: 0.7753 - acc: 0.7361\n",
            "Epoch 35/50\n",
            "5130/5130 [==============================] - 1s 222us/sample - loss: 0.7703 - acc: 0.7376\n",
            "Epoch 36/50\n",
            "5130/5130 [==============================] - 1s 220us/sample - loss: 0.7585 - acc: 0.7446\n",
            "Epoch 37/50\n",
            "5130/5130 [==============================] - 1s 221us/sample - loss: 0.7636 - acc: 0.7363\n",
            "Epoch 38/50\n",
            "5130/5130 [==============================] - 1s 222us/sample - loss: 0.7573 - acc: 0.7515\n",
            "Epoch 39/50\n",
            "5130/5130 [==============================] - 1s 216us/sample - loss: 0.7447 - acc: 0.7450\n",
            "Epoch 40/50\n",
            "5130/5130 [==============================] - 1s 226us/sample - loss: 0.7480 - acc: 0.7394\n",
            "Epoch 41/50\n",
            "5130/5130 [==============================] - 1s 225us/sample - loss: 0.7501 - acc: 0.7468\n",
            "Epoch 42/50\n",
            "5130/5130 [==============================] - 1s 217us/sample - loss: 0.7480 - acc: 0.7538\n",
            "Epoch 43/50\n",
            "5130/5130 [==============================] - 1s 218us/sample - loss: 0.7404 - acc: 0.7581\n",
            "Epoch 44/50\n",
            "5130/5130 [==============================] - 1s 215us/sample - loss: 0.7241 - acc: 0.7591\n",
            "Epoch 45/50\n",
            "5130/5130 [==============================] - 1s 211us/sample - loss: 0.7250 - acc: 0.7542\n",
            "Epoch 46/50\n",
            "5130/5130 [==============================] - 1s 210us/sample - loss: 0.7282 - acc: 0.7520\n",
            "Epoch 47/50\n",
            "5130/5130 [==============================] - 1s 213us/sample - loss: 0.7135 - acc: 0.7542\n",
            "Epoch 48/50\n",
            "5130/5130 [==============================] - 1s 210us/sample - loss: 0.7139 - acc: 0.7563\n",
            "Epoch 49/50\n",
            "5130/5130 [==============================] - 1s 210us/sample - loss: 0.7095 - acc: 0.7569\n",
            "Epoch 50/50\n",
            "5130/5130 [==============================] - 1s 212us/sample - loss: 0.7145 - acc: 0.7628\n",
            "Epoch 1/50\n",
            "5130/5130 [==============================] - 1s 262us/sample - loss: 1.6648 - acc: 0.4125\n",
            "Epoch 2/50\n",
            "5130/5130 [==============================] - 1s 236us/sample - loss: 1.3528 - acc: 0.5263\n",
            "Epoch 3/50\n",
            "5130/5130 [==============================] - 1s 236us/sample - loss: 1.2577 - acc: 0.5550\n",
            "Epoch 4/50\n",
            "5130/5130 [==============================] - 1s 236us/sample - loss: 1.1896 - acc: 0.5852\n",
            "Epoch 5/50\n",
            "5130/5130 [==============================] - 1s 233us/sample - loss: 1.1418 - acc: 0.5965\n",
            "Epoch 6/50\n",
            "5130/5130 [==============================] - 1s 238us/sample - loss: 1.1011 - acc: 0.6103\n",
            "Epoch 7/50\n",
            "5130/5130 [==============================] - 1s 238us/sample - loss: 1.0738 - acc: 0.6177\n",
            "Epoch 8/50\n",
            "5130/5130 [==============================] - 1s 236us/sample - loss: 1.0315 - acc: 0.6343\n",
            "Epoch 9/50\n",
            "5130/5130 [==============================] - 1s 236us/sample - loss: 1.0100 - acc: 0.6411\n",
            "Epoch 10/50\n",
            "5130/5130 [==============================] - 1s 235us/sample - loss: 0.9945 - acc: 0.6495\n",
            "Epoch 11/50\n",
            "5130/5130 [==============================] - 1s 233us/sample - loss: 0.9665 - acc: 0.6550\n",
            "Epoch 12/50\n",
            "5130/5130 [==============================] - 1s 232us/sample - loss: 0.9532 - acc: 0.6694\n",
            "Epoch 13/50\n",
            "5130/5130 [==============================] - 1s 236us/sample - loss: 0.9308 - acc: 0.6789\n",
            "Epoch 14/50\n",
            "5130/5130 [==============================] - 1s 235us/sample - loss: 0.9155 - acc: 0.6661\n",
            "Epoch 15/50\n",
            "5130/5130 [==============================] - 1s 236us/sample - loss: 0.8889 - acc: 0.6916\n",
            "Epoch 16/50\n",
            "5130/5130 [==============================] - 1s 235us/sample - loss: 0.8800 - acc: 0.6881\n",
            "Epoch 17/50\n",
            "5130/5130 [==============================] - 1s 237us/sample - loss: 0.8799 - acc: 0.6982\n",
            "Epoch 18/50\n",
            "5130/5130 [==============================] - 1s 229us/sample - loss: 0.8704 - acc: 0.6938\n",
            "Epoch 19/50\n",
            "5130/5130 [==============================] - 1s 235us/sample - loss: 0.8604 - acc: 0.6975\n",
            "Epoch 20/50\n",
            "5130/5130 [==============================] - 1s 232us/sample - loss: 0.8448 - acc: 0.7027\n",
            "Epoch 21/50\n",
            "5130/5130 [==============================] - 1s 232us/sample - loss: 0.8364 - acc: 0.7094\n",
            "Epoch 22/50\n",
            "5130/5130 [==============================] - 1s 231us/sample - loss: 0.8281 - acc: 0.7088\n",
            "Epoch 23/50\n",
            "5130/5130 [==============================] - 1s 230us/sample - loss: 0.8216 - acc: 0.7047\n",
            "Epoch 24/50\n",
            "5130/5130 [==============================] - 1s 229us/sample - loss: 0.8100 - acc: 0.7131\n",
            "Epoch 25/50\n",
            "5130/5130 [==============================] - 1s 229us/sample - loss: 0.7975 - acc: 0.7185\n",
            "Epoch 26/50\n",
            "5130/5130 [==============================] - 1s 226us/sample - loss: 0.7967 - acc: 0.7209\n",
            "Epoch 27/50\n",
            "5130/5130 [==============================] - 1s 230us/sample - loss: 0.7739 - acc: 0.7236\n",
            "Epoch 28/50\n",
            "5130/5130 [==============================] - 1s 227us/sample - loss: 0.7789 - acc: 0.7238\n",
            "Epoch 29/50\n",
            "5130/5130 [==============================] - 1s 225us/sample - loss: 0.7654 - acc: 0.7302\n",
            "Epoch 30/50\n",
            "5130/5130 [==============================] - 1s 231us/sample - loss: 0.7614 - acc: 0.7322\n",
            "Epoch 31/50\n",
            "5130/5130 [==============================] - 1s 226us/sample - loss: 0.7484 - acc: 0.7376\n",
            "Epoch 32/50\n",
            "5130/5130 [==============================] - 1s 227us/sample - loss: 0.7531 - acc: 0.7351\n",
            "Epoch 33/50\n",
            "5130/5130 [==============================] - 1s 230us/sample - loss: 0.7458 - acc: 0.7407\n",
            "Epoch 34/50\n",
            "5130/5130 [==============================] - 1s 224us/sample - loss: 0.7353 - acc: 0.7462\n",
            "Epoch 35/50\n",
            "5130/5130 [==============================] - 1s 226us/sample - loss: 0.7582 - acc: 0.7318\n",
            "Epoch 36/50\n",
            "5130/5130 [==============================] - 1s 224us/sample - loss: 0.7174 - acc: 0.7466\n",
            "Epoch 37/50\n",
            "5130/5130 [==============================] - 1s 227us/sample - loss: 0.7205 - acc: 0.7421\n",
            "Epoch 38/50\n",
            "5130/5130 [==============================] - 1s 226us/sample - loss: 0.7208 - acc: 0.7423\n",
            "Epoch 39/50\n",
            "5130/5130 [==============================] - 1s 227us/sample - loss: 0.7266 - acc: 0.7419\n",
            "Epoch 40/50\n",
            "5130/5130 [==============================] - 1s 233us/sample - loss: 0.7083 - acc: 0.7493\n",
            "Epoch 41/50\n",
            "5130/5130 [==============================] - 1s 226us/sample - loss: 0.7017 - acc: 0.7598\n",
            "Epoch 42/50\n",
            "5130/5130 [==============================] - 1s 225us/sample - loss: 0.6958 - acc: 0.7575\n",
            "Epoch 43/50\n",
            "5130/5130 [==============================] - 1s 230us/sample - loss: 0.7044 - acc: 0.7466\n",
            "Epoch 44/50\n",
            "5130/5130 [==============================] - 1s 231us/sample - loss: 0.6827 - acc: 0.7641\n",
            "Epoch 45/50\n",
            "5130/5130 [==============================] - 1s 229us/sample - loss: 0.6723 - acc: 0.7581\n",
            "Epoch 46/50\n",
            "5130/5130 [==============================] - 1s 227us/sample - loss: 0.6808 - acc: 0.7608\n",
            "Epoch 47/50\n",
            "5130/5130 [==============================] - 1s 229us/sample - loss: 0.6720 - acc: 0.7569\n",
            "Epoch 48/50\n",
            "5130/5130 [==============================] - 1s 232us/sample - loss: 0.6704 - acc: 0.7596\n",
            "Epoch 49/50\n",
            "5130/5130 [==============================] - 1s 226us/sample - loss: 0.6701 - acc: 0.7715\n",
            "Epoch 50/50\n",
            "5130/5130 [==============================] - 1s 228us/sample - loss: 0.6551 - acc: 0.7702\n",
            "Epoch 1/50\n",
            "5130/5130 [==============================] - 1s 235us/sample - loss: 1.6523 - acc: 0.4154\n",
            "Epoch 2/50\n",
            "5130/5130 [==============================] - 1s 207us/sample - loss: 1.3500 - acc: 0.5187\n",
            "Epoch 3/50\n",
            "5130/5130 [==============================] - 1s 199us/sample - loss: 1.2600 - acc: 0.5542\n",
            "Epoch 4/50\n",
            "5130/5130 [==============================] - 1s 222us/sample - loss: 1.1895 - acc: 0.5780\n",
            "Epoch 5/50\n",
            "5130/5130 [==============================] - 1s 219us/sample - loss: 1.1462 - acc: 0.5936\n",
            "Epoch 6/50\n",
            "5130/5130 [==============================] - 1s 212us/sample - loss: 1.1012 - acc: 0.6135\n",
            "Epoch 7/50\n",
            "5130/5130 [==============================] - 1s 214us/sample - loss: 1.0683 - acc: 0.6228\n",
            "Epoch 8/50\n",
            "5130/5130 [==============================] - 1s 213us/sample - loss: 1.0514 - acc: 0.6251\n",
            "Epoch 9/50\n",
            "5130/5130 [==============================] - 1s 210us/sample - loss: 1.0314 - acc: 0.6281\n",
            "Epoch 10/50\n",
            "5130/5130 [==============================] - 1s 215us/sample - loss: 1.0023 - acc: 0.6464\n",
            "Epoch 11/50\n",
            "5130/5130 [==============================] - 1s 209us/sample - loss: 0.9875 - acc: 0.6480\n",
            "Epoch 12/50\n",
            "5130/5130 [==============================] - 1s 214us/sample - loss: 0.9717 - acc: 0.6577\n",
            "Epoch 13/50\n",
            "5130/5130 [==============================] - 1s 209us/sample - loss: 0.9597 - acc: 0.6641\n",
            "Epoch 14/50\n",
            "5130/5130 [==============================] - 1s 210us/sample - loss: 0.9297 - acc: 0.6758\n",
            "Epoch 15/50\n",
            "5130/5130 [==============================] - 1s 213us/sample - loss: 0.9337 - acc: 0.6766\n",
            "Epoch 16/50\n",
            "5130/5130 [==============================] - 1s 208us/sample - loss: 0.9127 - acc: 0.6793\n",
            "Epoch 17/50\n",
            "5130/5130 [==============================] - 1s 215us/sample - loss: 0.9129 - acc: 0.6832\n",
            "Epoch 18/50\n",
            "5130/5130 [==============================] - 1s 211us/sample - loss: 0.9021 - acc: 0.6963\n",
            "Epoch 19/50\n",
            "5130/5130 [==============================] - 1s 209us/sample - loss: 0.8953 - acc: 0.6959\n",
            "Epoch 20/50\n",
            "5130/5130 [==============================] - 1s 208us/sample - loss: 0.8788 - acc: 0.6943\n",
            "Epoch 21/50\n",
            "5130/5130 [==============================] - 1s 210us/sample - loss: 0.8632 - acc: 0.6977\n",
            "Epoch 22/50\n",
            "5130/5130 [==============================] - 1s 208us/sample - loss: 0.8662 - acc: 0.6979\n",
            "Epoch 23/50\n",
            "5130/5130 [==============================] - 1s 209us/sample - loss: 0.8442 - acc: 0.7119\n",
            "Epoch 24/50\n",
            "5130/5130 [==============================] - 1s 211us/sample - loss: 0.8439 - acc: 0.7107\n",
            "Epoch 25/50\n",
            "5130/5130 [==============================] - 1s 210us/sample - loss: 0.8503 - acc: 0.6967\n",
            "Epoch 26/50\n",
            "5130/5130 [==============================] - 1s 210us/sample - loss: 0.8357 - acc: 0.7136\n",
            "Epoch 27/50\n",
            "5130/5130 [==============================] - 1s 211us/sample - loss: 0.8207 - acc: 0.7136\n",
            "Epoch 28/50\n",
            "5130/5130 [==============================] - 1s 210us/sample - loss: 0.8293 - acc: 0.7175\n",
            "Epoch 29/50\n",
            "5130/5130 [==============================] - 1s 212us/sample - loss: 0.8093 - acc: 0.7250\n",
            "Epoch 30/50\n",
            "5130/5130 [==============================] - 1s 209us/sample - loss: 0.8201 - acc: 0.7238\n",
            "Epoch 31/50\n",
            "5130/5130 [==============================] - 1s 209us/sample - loss: 0.7855 - acc: 0.7312\n",
            "Epoch 32/50\n",
            "5130/5130 [==============================] - 1s 210us/sample - loss: 0.8035 - acc: 0.7263\n",
            "Epoch 33/50\n",
            "5130/5130 [==============================] - 1s 214us/sample - loss: 0.7883 - acc: 0.7327\n",
            "Epoch 34/50\n",
            "5130/5130 [==============================] - 1s 211us/sample - loss: 0.7809 - acc: 0.7372\n",
            "Epoch 35/50\n",
            "5130/5130 [==============================] - 1s 218us/sample - loss: 0.7624 - acc: 0.7392\n",
            "Epoch 36/50\n",
            "5130/5130 [==============================] - 1s 217us/sample - loss: 0.7656 - acc: 0.7421\n",
            "Epoch 37/50\n",
            "5130/5130 [==============================] - 1s 210us/sample - loss: 0.7692 - acc: 0.7415\n",
            "Epoch 38/50\n",
            "5130/5130 [==============================] - 1s 216us/sample - loss: 0.7659 - acc: 0.7444\n",
            "Epoch 39/50\n",
            "5130/5130 [==============================] - 1s 214us/sample - loss: 0.7559 - acc: 0.7425\n",
            "Epoch 40/50\n",
            "5130/5130 [==============================] - 1s 215us/sample - loss: 0.7584 - acc: 0.7363\n",
            "Epoch 41/50\n",
            "5130/5130 [==============================] - 1s 212us/sample - loss: 0.7537 - acc: 0.7495\n",
            "Epoch 42/50\n",
            "5130/5130 [==============================] - 1s 211us/sample - loss: 0.7452 - acc: 0.7480\n",
            "Epoch 43/50\n",
            "5130/5130 [==============================] - 1s 213us/sample - loss: 0.7448 - acc: 0.7499\n",
            "Epoch 44/50\n",
            "5130/5130 [==============================] - 1s 218us/sample - loss: 0.7426 - acc: 0.7526\n",
            "Epoch 45/50\n",
            "5130/5130 [==============================] - 1s 216us/sample - loss: 0.7097 - acc: 0.7602\n",
            "Epoch 46/50\n",
            "5130/5130 [==============================] - 1s 217us/sample - loss: 0.7228 - acc: 0.7595\n",
            "Epoch 47/50\n",
            "5130/5130 [==============================] - 1s 216us/sample - loss: 0.7438 - acc: 0.7542\n",
            "Epoch 48/50\n",
            "5130/5130 [==============================] - 1s 215us/sample - loss: 0.7282 - acc: 0.7530\n",
            "Epoch 49/50\n",
            "5130/5130 [==============================] - 1s 217us/sample - loss: 0.7416 - acc: 0.7534\n",
            "Epoch 50/50\n",
            "5130/5130 [==============================] - 1s 216us/sample - loss: 0.7261 - acc: 0.7598\n",
            "Epoch 1/50\n",
            "5130/5130 [==============================] - 1s 269us/sample - loss: 1.6538 - acc: 0.4179\n",
            "Epoch 2/50\n",
            "5130/5130 [==============================] - 1s 237us/sample - loss: 1.3410 - acc: 0.5279\n",
            "Epoch 3/50\n",
            "5130/5130 [==============================] - 1s 239us/sample - loss: 1.2472 - acc: 0.5561\n",
            "Epoch 4/50\n",
            "5130/5130 [==============================] - 1s 236us/sample - loss: 1.1897 - acc: 0.5774\n",
            "Epoch 5/50\n",
            "5130/5130 [==============================] - 1s 237us/sample - loss: 1.1487 - acc: 0.5912\n",
            "Epoch 6/50\n",
            "5130/5130 [==============================] - 1s 233us/sample - loss: 1.0929 - acc: 0.6090\n",
            "Epoch 7/50\n",
            "5130/5130 [==============================] - 1s 242us/sample - loss: 1.0678 - acc: 0.6168\n",
            "Epoch 8/50\n",
            "5130/5130 [==============================] - 1s 244us/sample - loss: 1.0366 - acc: 0.6318\n",
            "Epoch 9/50\n",
            "5130/5130 [==============================] - 1s 231us/sample - loss: 1.0090 - acc: 0.6366\n",
            "Epoch 10/50\n",
            "5130/5130 [==============================] - 1s 242us/sample - loss: 0.9845 - acc: 0.6505\n",
            "Epoch 11/50\n",
            "5130/5130 [==============================] - 1s 237us/sample - loss: 0.9775 - acc: 0.6589\n",
            "Epoch 12/50\n",
            "5130/5130 [==============================] - 1s 247us/sample - loss: 0.9445 - acc: 0.6690\n",
            "Epoch 13/50\n",
            "5130/5130 [==============================] - 1s 246us/sample - loss: 0.9346 - acc: 0.6750\n",
            "Epoch 14/50\n",
            "5130/5130 [==============================] - 1s 235us/sample - loss: 0.9157 - acc: 0.6754\n",
            "Epoch 15/50\n",
            "5130/5130 [==============================] - 1s 239us/sample - loss: 0.9000 - acc: 0.6815\n",
            "Epoch 16/50\n",
            "5130/5130 [==============================] - 1s 238us/sample - loss: 0.9006 - acc: 0.6772\n",
            "Epoch 17/50\n",
            "5130/5130 [==============================] - 1s 239us/sample - loss: 0.8905 - acc: 0.6848\n",
            "Epoch 18/50\n",
            "5130/5130 [==============================] - 1s 236us/sample - loss: 0.8744 - acc: 0.6924\n",
            "Epoch 19/50\n",
            "5130/5130 [==============================] - 1s 234us/sample - loss: 0.8581 - acc: 0.7057\n",
            "Epoch 20/50\n",
            "5130/5130 [==============================] - 1s 237us/sample - loss: 0.8503 - acc: 0.7039\n",
            "Epoch 21/50\n",
            "5130/5130 [==============================] - 1s 249us/sample - loss: 0.8481 - acc: 0.7031\n",
            "Epoch 22/50\n",
            "5130/5130 [==============================] - 1s 248us/sample - loss: 0.8276 - acc: 0.7135\n",
            "Epoch 23/50\n",
            "5130/5130 [==============================] - 1s 238us/sample - loss: 0.8233 - acc: 0.7101\n",
            "Epoch 24/50\n",
            "5130/5130 [==============================] - 1s 235us/sample - loss: 0.8179 - acc: 0.7164\n",
            "Epoch 25/50\n",
            "5130/5130 [==============================] - 1s 235us/sample - loss: 0.8164 - acc: 0.7103\n",
            "Epoch 26/50\n",
            "5130/5130 [==============================] - 1s 230us/sample - loss: 0.7918 - acc: 0.7201\n",
            "Epoch 27/50\n",
            "5130/5130 [==============================] - 1s 232us/sample - loss: 0.7921 - acc: 0.7240\n",
            "Epoch 28/50\n",
            "5130/5130 [==============================] - 1s 230us/sample - loss: 0.7826 - acc: 0.7228\n",
            "Epoch 29/50\n",
            "5130/5130 [==============================] - 1s 230us/sample - loss: 0.7836 - acc: 0.7234\n",
            "Epoch 30/50\n",
            "5130/5130 [==============================] - 1s 230us/sample - loss: 0.7701 - acc: 0.7380\n",
            "Epoch 31/50\n",
            "5130/5130 [==============================] - 1s 228us/sample - loss: 0.7623 - acc: 0.7304\n",
            "Epoch 32/50\n",
            "5130/5130 [==============================] - 1s 243us/sample - loss: 0.7677 - acc: 0.7326\n",
            "Epoch 33/50\n",
            "5130/5130 [==============================] - 1s 258us/sample - loss: 0.7386 - acc: 0.7427\n",
            "Epoch 34/50\n",
            "5130/5130 [==============================] - 1s 256us/sample - loss: 0.7357 - acc: 0.7429\n",
            "Epoch 35/50\n",
            "5130/5130 [==============================] - 1s 258us/sample - loss: 0.7385 - acc: 0.7409\n",
            "Epoch 36/50\n",
            "5130/5130 [==============================] - 1s 259us/sample - loss: 0.7264 - acc: 0.7464\n",
            "Epoch 37/50\n",
            "5130/5130 [==============================] - 1s 253us/sample - loss: 0.7251 - acc: 0.7417\n",
            "Epoch 38/50\n",
            "5130/5130 [==============================] - 1s 257us/sample - loss: 0.7234 - acc: 0.7517\n",
            "Epoch 39/50\n",
            "5130/5130 [==============================] - 1s 256us/sample - loss: 0.7117 - acc: 0.7481\n",
            "Epoch 40/50\n",
            "5130/5130 [==============================] - 1s 248us/sample - loss: 0.7143 - acc: 0.7501\n",
            "Epoch 41/50\n",
            "5130/5130 [==============================] - 1s 234us/sample - loss: 0.6973 - acc: 0.7577\n",
            "Epoch 42/50\n",
            "5130/5130 [==============================] - 1s 231us/sample - loss: 0.7088 - acc: 0.7485\n",
            "Epoch 43/50\n",
            "5130/5130 [==============================] - 1s 233us/sample - loss: 0.6993 - acc: 0.7456\n",
            "Epoch 44/50\n",
            "5130/5130 [==============================] - 1s 237us/sample - loss: 0.6819 - acc: 0.7643\n",
            "Epoch 45/50\n",
            "5130/5130 [==============================] - 1s 235us/sample - loss: 0.6918 - acc: 0.7589\n",
            "Epoch 46/50\n",
            "5130/5130 [==============================] - 1s 236us/sample - loss: 0.6802 - acc: 0.7630\n",
            "Epoch 47/50\n",
            "5130/5130 [==============================] - 1s 234us/sample - loss: 0.6686 - acc: 0.7673\n",
            "Epoch 48/50\n",
            "5130/5130 [==============================] - 1s 235us/sample - loss: 0.6706 - acc: 0.7583\n",
            "Epoch 49/50\n",
            "5130/5130 [==============================] - 1s 240us/sample - loss: 0.6705 - acc: 0.7678\n",
            "Epoch 50/50\n",
            "5130/5130 [==============================] - 1s 237us/sample - loss: 0.6668 - acc: 0.7686\n",
            "Epoch 1/50\n",
            "5130/5130 [==============================] - 1s 246us/sample - loss: 1.6364 - acc: 0.4281\n",
            "Epoch 2/50\n",
            "5130/5130 [==============================] - 1s 216us/sample - loss: 1.3355 - acc: 0.5253\n",
            "Epoch 3/50\n",
            "5130/5130 [==============================] - 1s 215us/sample - loss: 1.2517 - acc: 0.5550\n",
            "Epoch 4/50\n",
            "5130/5130 [==============================] - 1s 216us/sample - loss: 1.1922 - acc: 0.5729\n",
            "Epoch 5/50\n",
            "5130/5130 [==============================] - 1s 216us/sample - loss: 1.1377 - acc: 0.5984\n",
            "Epoch 6/50\n",
            "5130/5130 [==============================] - 1s 215us/sample - loss: 1.1089 - acc: 0.6041\n",
            "Epoch 7/50\n",
            "5130/5130 [==============================] - 1s 215us/sample - loss: 1.0761 - acc: 0.6197\n",
            "Epoch 8/50\n",
            "5130/5130 [==============================] - 1s 214us/sample - loss: 1.0599 - acc: 0.6265\n",
            "Epoch 9/50\n",
            "5130/5130 [==============================] - 1s 214us/sample - loss: 1.0200 - acc: 0.6464\n",
            "Epoch 10/50\n",
            "5130/5130 [==============================] - 1s 214us/sample - loss: 1.0013 - acc: 0.6464\n",
            "Epoch 11/50\n",
            "5130/5130 [==============================] - 1s 213us/sample - loss: 0.9916 - acc: 0.6520\n",
            "Epoch 12/50\n",
            "5130/5130 [==============================] - 1s 217us/sample - loss: 0.9719 - acc: 0.6653\n",
            "Epoch 13/50\n",
            "5130/5130 [==============================] - 1s 215us/sample - loss: 0.9554 - acc: 0.6663\n",
            "Epoch 14/50\n",
            "5130/5130 [==============================] - 1s 219us/sample - loss: 0.9457 - acc: 0.6725\n",
            "Epoch 15/50\n",
            "5130/5130 [==============================] - 1s 216us/sample - loss: 0.9237 - acc: 0.6764\n",
            "Epoch 16/50\n",
            "5130/5130 [==============================] - 1s 217us/sample - loss: 0.9128 - acc: 0.6840\n",
            "Epoch 17/50\n",
            "5130/5130 [==============================] - 1s 216us/sample - loss: 0.9206 - acc: 0.6729\n",
            "Epoch 18/50\n",
            "5130/5130 [==============================] - 1s 216us/sample - loss: 0.9005 - acc: 0.6893\n",
            "Epoch 19/50\n",
            "5130/5130 [==============================] - 1s 217us/sample - loss: 0.8874 - acc: 0.6918\n",
            "Epoch 20/50\n",
            "5130/5130 [==============================] - 1s 218us/sample - loss: 0.8777 - acc: 0.6953\n",
            "Epoch 21/50\n",
            "5130/5130 [==============================] - 1s 215us/sample - loss: 0.8572 - acc: 0.6973\n",
            "Epoch 22/50\n",
            "5130/5130 [==============================] - 1s 215us/sample - loss: 0.8579 - acc: 0.7027\n",
            "Epoch 23/50\n",
            "5130/5130 [==============================] - 1s 215us/sample - loss: 0.8428 - acc: 0.7115\n",
            "Epoch 24/50\n",
            "5130/5130 [==============================] - 1s 214us/sample - loss: 0.8366 - acc: 0.7138\n",
            "Epoch 25/50\n",
            "5130/5130 [==============================] - 1s 213us/sample - loss: 0.8527 - acc: 0.7070\n",
            "Epoch 26/50\n",
            "5130/5130 [==============================] - 1s 215us/sample - loss: 0.8207 - acc: 0.7216\n",
            "Epoch 27/50\n",
            "5130/5130 [==============================] - 1s 214us/sample - loss: 0.8231 - acc: 0.7175\n",
            "Epoch 28/50\n",
            "5130/5130 [==============================] - 1s 212us/sample - loss: 0.8231 - acc: 0.7205\n",
            "Epoch 29/50\n",
            "5130/5130 [==============================] - 1s 215us/sample - loss: 0.8038 - acc: 0.7195\n",
            "Epoch 30/50\n",
            "5130/5130 [==============================] - 1s 217us/sample - loss: 0.7985 - acc: 0.7232\n",
            "Epoch 31/50\n",
            "5130/5130 [==============================] - 1s 216us/sample - loss: 0.8042 - acc: 0.7271\n",
            "Epoch 32/50\n",
            "5130/5130 [==============================] - 1s 231us/sample - loss: 0.7860 - acc: 0.7312\n",
            "Epoch 33/50\n",
            "5130/5130 [==============================] - 1s 233us/sample - loss: 0.7890 - acc: 0.7292\n",
            "Epoch 34/50\n",
            "5130/5130 [==============================] - 1s 231us/sample - loss: 0.7736 - acc: 0.7437\n",
            "Epoch 35/50\n",
            "5130/5130 [==============================] - 1s 233us/sample - loss: 0.7678 - acc: 0.7409\n",
            "Epoch 36/50\n",
            "5130/5130 [==============================] - 1s 239us/sample - loss: 0.7666 - acc: 0.7435\n",
            "Epoch 37/50\n",
            "5130/5130 [==============================] - 1s 226us/sample - loss: 0.7691 - acc: 0.7417\n",
            "Epoch 38/50\n",
            "5130/5130 [==============================] - 1s 217us/sample - loss: 0.7643 - acc: 0.7349\n",
            "Epoch 39/50\n",
            "5130/5130 [==============================] - 1s 224us/sample - loss: 0.7439 - acc: 0.7530\n",
            "Epoch 40/50\n",
            "5130/5130 [==============================] - 1s 218us/sample - loss: 0.7485 - acc: 0.7517\n",
            "Epoch 41/50\n",
            "5130/5130 [==============================] - 1s 220us/sample - loss: 0.7527 - acc: 0.7466\n",
            "Epoch 42/50\n",
            "5130/5130 [==============================] - 1s 220us/sample - loss: 0.7239 - acc: 0.7534\n",
            "Epoch 43/50\n",
            "5130/5130 [==============================] - 1s 217us/sample - loss: 0.7448 - acc: 0.7476\n",
            "Epoch 44/50\n",
            "5130/5130 [==============================] - 1s 216us/sample - loss: 0.7405 - acc: 0.7497\n",
            "Epoch 45/50\n",
            "5130/5130 [==============================] - 1s 219us/sample - loss: 0.7418 - acc: 0.7509\n",
            "Epoch 46/50\n",
            "5130/5130 [==============================] - 1s 216us/sample - loss: 0.7329 - acc: 0.7511\n",
            "Epoch 47/50\n",
            "5130/5130 [==============================] - 1s 218us/sample - loss: 0.7139 - acc: 0.7641\n",
            "Epoch 48/50\n",
            "5130/5130 [==============================] - 1s 217us/sample - loss: 0.7217 - acc: 0.7602\n",
            "Epoch 49/50\n",
            "5130/5130 [==============================] - 1s 215us/sample - loss: 0.7067 - acc: 0.7626\n",
            "Epoch 50/50\n",
            "5130/5130 [==============================] - 1s 215us/sample - loss: 0.7145 - acc: 0.7678\n",
            "Epoch 1/50\n",
            "5130/5130 [==============================] - 1s 268us/sample - loss: 1.6287 - acc: 0.4314\n",
            "Epoch 2/50\n",
            "5130/5130 [==============================] - 1s 237us/sample - loss: 1.3358 - acc: 0.5230\n",
            "Epoch 3/50\n",
            "5130/5130 [==============================] - 1s 234us/sample - loss: 1.2533 - acc: 0.5538\n",
            "Epoch 4/50\n",
            "5130/5130 [==============================] - 1s 233us/sample - loss: 1.1706 - acc: 0.5899\n",
            "Epoch 5/50\n",
            "5130/5130 [==============================] - 1s 232us/sample - loss: 1.1315 - acc: 0.5959\n",
            "Epoch 6/50\n",
            "5130/5130 [==============================] - 1s 235us/sample - loss: 1.1000 - acc: 0.6080\n",
            "Epoch 7/50\n",
            "5130/5130 [==============================] - 1s 235us/sample - loss: 1.0734 - acc: 0.6154\n",
            "Epoch 8/50\n",
            "5130/5130 [==============================] - 1s 238us/sample - loss: 1.0418 - acc: 0.6271\n",
            "Epoch 9/50\n",
            "5130/5130 [==============================] - 1s 238us/sample - loss: 1.0138 - acc: 0.6419\n",
            "Epoch 10/50\n",
            "5130/5130 [==============================] - 1s 235us/sample - loss: 0.9880 - acc: 0.6501\n",
            "Epoch 11/50\n",
            "5130/5130 [==============================] - 1s 236us/sample - loss: 0.9616 - acc: 0.6511\n",
            "Epoch 12/50\n",
            "5130/5130 [==============================] - 1s 238us/sample - loss: 0.9521 - acc: 0.6635\n",
            "Epoch 13/50\n",
            "5130/5130 [==============================] - 1s 240us/sample - loss: 0.9374 - acc: 0.6762\n",
            "Epoch 14/50\n",
            "5130/5130 [==============================] - 1s 239us/sample - loss: 0.9154 - acc: 0.6772\n",
            "Epoch 15/50\n",
            "5130/5130 [==============================] - 1s 240us/sample - loss: 0.9065 - acc: 0.6793\n",
            "Epoch 16/50\n",
            "5130/5130 [==============================] - 1s 235us/sample - loss: 0.8907 - acc: 0.6815\n",
            "Epoch 17/50\n",
            "5130/5130 [==============================] - 1s 235us/sample - loss: 0.8738 - acc: 0.6916\n",
            "Epoch 18/50\n",
            "5130/5130 [==============================] - 1s 235us/sample - loss: 0.8659 - acc: 0.6904\n",
            "Epoch 19/50\n",
            "5130/5130 [==============================] - 1s 236us/sample - loss: 0.8629 - acc: 0.6957\n",
            "Epoch 20/50\n",
            "5130/5130 [==============================] - 1s 237us/sample - loss: 0.8531 - acc: 0.7045\n",
            "Epoch 21/50\n",
            "5130/5130 [==============================] - 1s 235us/sample - loss: 0.8372 - acc: 0.7103\n",
            "Epoch 22/50\n",
            "5130/5130 [==============================] - 1s 237us/sample - loss: 0.8374 - acc: 0.7029\n",
            "Epoch 23/50\n",
            "5130/5130 [==============================] - 1s 236us/sample - loss: 0.8231 - acc: 0.7072\n",
            "Epoch 24/50\n",
            "5130/5130 [==============================] - 1s 236us/sample - loss: 0.8158 - acc: 0.7109\n",
            "Epoch 25/50\n",
            "5130/5130 [==============================] - 1s 235us/sample - loss: 0.7992 - acc: 0.7173\n",
            "Epoch 26/50\n",
            "5130/5130 [==============================] - 1s 239us/sample - loss: 0.7906 - acc: 0.7228\n",
            "Epoch 27/50\n",
            "5130/5130 [==============================] - 1s 237us/sample - loss: 0.7842 - acc: 0.7242\n",
            "Epoch 28/50\n",
            "5130/5130 [==============================] - 1s 234us/sample - loss: 0.7833 - acc: 0.7331\n",
            "Epoch 29/50\n",
            "5130/5130 [==============================] - 1s 233us/sample - loss: 0.7802 - acc: 0.7197\n",
            "Epoch 30/50\n",
            "5130/5130 [==============================] - 1s 236us/sample - loss: 0.7770 - acc: 0.7287\n",
            "Epoch 31/50\n",
            "5130/5130 [==============================] - 1s 236us/sample - loss: 0.7654 - acc: 0.7357\n",
            "Epoch 32/50\n",
            "5130/5130 [==============================] - 1s 236us/sample - loss: 0.7549 - acc: 0.7417\n",
            "Epoch 33/50\n",
            "5130/5130 [==============================] - 1s 237us/sample - loss: 0.7482 - acc: 0.7444\n",
            "Epoch 34/50\n",
            "5130/5130 [==============================] - 1s 239us/sample - loss: 0.7407 - acc: 0.7365\n",
            "Epoch 35/50\n",
            "5130/5130 [==============================] - 1s 239us/sample - loss: 0.7357 - acc: 0.7481\n",
            "Epoch 36/50\n",
            "5130/5130 [==============================] - 1s 236us/sample - loss: 0.7400 - acc: 0.7388\n",
            "Epoch 37/50\n",
            "5130/5130 [==============================] - 1s 236us/sample - loss: 0.7194 - acc: 0.7448\n",
            "Epoch 38/50\n",
            "5130/5130 [==============================] - 1s 236us/sample - loss: 0.7144 - acc: 0.7534\n",
            "Epoch 39/50\n",
            "5130/5130 [==============================] - 1s 237us/sample - loss: 0.7237 - acc: 0.7427\n",
            "Epoch 40/50\n",
            "5130/5130 [==============================] - 1s 239us/sample - loss: 0.7093 - acc: 0.7540\n",
            "Epoch 41/50\n",
            "5130/5130 [==============================] - 1s 236us/sample - loss: 0.7064 - acc: 0.7563\n",
            "Epoch 42/50\n",
            "5130/5130 [==============================] - 1s 236us/sample - loss: 0.7047 - acc: 0.7501\n",
            "Epoch 43/50\n",
            "5130/5130 [==============================] - 1s 239us/sample - loss: 0.6966 - acc: 0.7567\n",
            "Epoch 44/50\n",
            "5130/5130 [==============================] - 1s 237us/sample - loss: 0.6932 - acc: 0.7558\n",
            "Epoch 45/50\n",
            "5130/5130 [==============================] - 1s 241us/sample - loss: 0.6822 - acc: 0.7610\n",
            "Epoch 46/50\n",
            "5130/5130 [==============================] - 1s 233us/sample - loss: 0.6731 - acc: 0.7628\n",
            "Epoch 47/50\n",
            "5130/5130 [==============================] - 1s 230us/sample - loss: 0.6672 - acc: 0.7661\n",
            "Epoch 48/50\n",
            "5130/5130 [==============================] - 1s 235us/sample - loss: 0.6705 - acc: 0.7614\n",
            "Epoch 49/50\n",
            "5130/5130 [==============================] - 1s 237us/sample - loss: 0.6732 - acc: 0.7602\n",
            "Epoch 50/50\n",
            "5130/5130 [==============================] - 1s 251us/sample - loss: 0.6782 - acc: 0.7604\n",
            "Epoch 1/50\n",
            "5130/5130 [==============================] - 1s 144us/sample - loss: 1.7249 - acc: 0.4027\n",
            "Epoch 2/50\n",
            "5130/5130 [==============================] - 1s 108us/sample - loss: 1.4135 - acc: 0.5027\n",
            "Epoch 3/50\n",
            "5130/5130 [==============================] - 1s 108us/sample - loss: 1.3015 - acc: 0.5433\n",
            "Epoch 4/50\n",
            "5130/5130 [==============================] - 1s 110us/sample - loss: 1.2396 - acc: 0.5559\n",
            "Epoch 5/50\n",
            "5130/5130 [==============================] - 1s 108us/sample - loss: 1.1949 - acc: 0.5842\n",
            "Epoch 6/50\n",
            "5130/5130 [==============================] - 1s 111us/sample - loss: 1.1666 - acc: 0.5815\n",
            "Epoch 7/50\n",
            "5130/5130 [==============================] - 1s 109us/sample - loss: 1.1252 - acc: 0.6060\n",
            "Epoch 8/50\n",
            "5130/5130 [==============================] - 1s 113us/sample - loss: 1.1026 - acc: 0.6191\n",
            "Epoch 9/50\n",
            "5130/5130 [==============================] - 1s 107us/sample - loss: 1.0809 - acc: 0.6205\n",
            "Epoch 10/50\n",
            "5130/5130 [==============================] - 1s 111us/sample - loss: 1.0513 - acc: 0.6248\n",
            "Epoch 11/50\n",
            "5130/5130 [==============================] - 1s 110us/sample - loss: 1.0275 - acc: 0.6415\n",
            "Epoch 12/50\n",
            "5130/5130 [==============================] - 1s 108us/sample - loss: 1.0066 - acc: 0.6483\n",
            "Epoch 13/50\n",
            "5130/5130 [==============================] - 1s 113us/sample - loss: 1.0053 - acc: 0.6495\n",
            "Epoch 14/50\n",
            "5130/5130 [==============================] - 1s 115us/sample - loss: 0.9923 - acc: 0.6542\n",
            "Epoch 15/50\n",
            "5130/5130 [==============================] - 1s 117us/sample - loss: 0.9639 - acc: 0.6620\n",
            "Epoch 16/50\n",
            "5130/5130 [==============================] - 1s 108us/sample - loss: 0.9545 - acc: 0.6651\n",
            "Epoch 17/50\n",
            "5130/5130 [==============================] - 1s 110us/sample - loss: 0.9494 - acc: 0.6676\n",
            "Epoch 18/50\n",
            "5130/5130 [==============================] - 1s 108us/sample - loss: 0.9442 - acc: 0.6768\n",
            "Epoch 19/50\n",
            "5130/5130 [==============================] - 1s 111us/sample - loss: 0.9335 - acc: 0.6807\n",
            "Epoch 20/50\n",
            "5130/5130 [==============================] - 1s 111us/sample - loss: 0.9151 - acc: 0.6836\n",
            "Epoch 21/50\n",
            "5130/5130 [==============================] - 1s 109us/sample - loss: 0.9076 - acc: 0.6805\n",
            "Epoch 22/50\n",
            "5130/5130 [==============================] - 1s 110us/sample - loss: 0.8980 - acc: 0.6871\n",
            "Epoch 23/50\n",
            "5130/5130 [==============================] - 1s 109us/sample - loss: 0.8881 - acc: 0.6858\n",
            "Epoch 24/50\n",
            "5130/5130 [==============================] - 1s 115us/sample - loss: 0.8837 - acc: 0.6895\n",
            "Epoch 25/50\n",
            "5130/5130 [==============================] - 1s 108us/sample - loss: 0.8755 - acc: 0.6981\n",
            "Epoch 26/50\n",
            "5130/5130 [==============================] - 1s 110us/sample - loss: 0.8635 - acc: 0.6982\n",
            "Epoch 27/50\n",
            "5130/5130 [==============================] - 1s 111us/sample - loss: 0.8549 - acc: 0.7039\n",
            "Epoch 28/50\n",
            "5130/5130 [==============================] - 1s 109us/sample - loss: 0.8575 - acc: 0.7004\n",
            "Epoch 29/50\n",
            "5130/5130 [==============================] - 1s 113us/sample - loss: 0.8306 - acc: 0.7131\n",
            "Epoch 30/50\n",
            "5130/5130 [==============================] - 1s 108us/sample - loss: 0.8296 - acc: 0.7121\n",
            "Epoch 31/50\n",
            "5130/5130 [==============================] - 1s 110us/sample - loss: 0.8328 - acc: 0.7092\n",
            "Epoch 32/50\n",
            "5130/5130 [==============================] - 1s 107us/sample - loss: 0.8172 - acc: 0.7154\n",
            "Epoch 33/50\n",
            "5130/5130 [==============================] - 1s 112us/sample - loss: 0.8087 - acc: 0.7211\n",
            "Epoch 34/50\n",
            "5130/5130 [==============================] - 1s 108us/sample - loss: 0.8093 - acc: 0.7131\n",
            "Epoch 35/50\n",
            "5130/5130 [==============================] - 1s 109us/sample - loss: 0.8069 - acc: 0.7232\n",
            "Epoch 36/50\n",
            "5130/5130 [==============================] - 1s 109us/sample - loss: 0.8086 - acc: 0.7234\n",
            "Epoch 37/50\n",
            "5130/5130 [==============================] - 1s 110us/sample - loss: 0.7915 - acc: 0.7279\n",
            "Epoch 38/50\n",
            "5130/5130 [==============================] - 1s 110us/sample - loss: 0.7957 - acc: 0.7255\n",
            "Epoch 39/50\n",
            "5130/5130 [==============================] - 1s 108us/sample - loss: 0.7767 - acc: 0.7318\n",
            "Epoch 40/50\n",
            "5130/5130 [==============================] - 1s 109us/sample - loss: 0.7715 - acc: 0.7345\n",
            "Epoch 41/50\n",
            "5130/5130 [==============================] - 1s 108us/sample - loss: 0.7872 - acc: 0.7292\n",
            "Epoch 42/50\n",
            "5130/5130 [==============================] - 1s 113us/sample - loss: 0.7779 - acc: 0.7285\n",
            "Epoch 43/50\n",
            "5130/5130 [==============================] - 1s 108us/sample - loss: 0.7517 - acc: 0.7361\n",
            "Epoch 44/50\n",
            "5130/5130 [==============================] - 1s 112us/sample - loss: 0.7614 - acc: 0.7439\n",
            "Epoch 45/50\n",
            "5130/5130 [==============================] - 1s 107us/sample - loss: 0.7631 - acc: 0.7329\n",
            "Epoch 46/50\n",
            "5130/5130 [==============================] - 1s 110us/sample - loss: 0.7466 - acc: 0.7439\n",
            "Epoch 47/50\n",
            "5130/5130 [==============================] - 1s 111us/sample - loss: 0.7267 - acc: 0.7470\n",
            "Epoch 48/50\n",
            "5130/5130 [==============================] - 1s 107us/sample - loss: 0.7538 - acc: 0.7431\n",
            "Epoch 49/50\n",
            "5130/5130 [==============================] - 1s 109us/sample - loss: 0.7376 - acc: 0.7442\n",
            "Epoch 50/50\n",
            "5130/5130 [==============================] - 1s 108us/sample - loss: 0.7482 - acc: 0.7429\n",
            "Epoch 1/50\n",
            "5130/5130 [==============================] - 1s 158us/sample - loss: 1.7557 - acc: 0.3786\n",
            "Epoch 2/50\n",
            "5130/5130 [==============================] - 1s 121us/sample - loss: 1.4085 - acc: 0.5133\n",
            "Epoch 3/50\n",
            "5130/5130 [==============================] - 1s 123us/sample - loss: 1.3075 - acc: 0.5355\n",
            "Epoch 4/50\n",
            "5130/5130 [==============================] - 1s 124us/sample - loss: 1.2396 - acc: 0.5593\n",
            "Epoch 5/50\n",
            "5130/5130 [==============================] - 1s 120us/sample - loss: 1.1919 - acc: 0.5788\n",
            "Epoch 6/50\n",
            "5130/5130 [==============================] - 1s 122us/sample - loss: 1.1591 - acc: 0.5875\n",
            "Epoch 7/50\n",
            "5130/5130 [==============================] - 1s 121us/sample - loss: 1.1229 - acc: 0.6029\n",
            "Epoch 8/50\n",
            "5130/5130 [==============================] - 1s 122us/sample - loss: 1.0953 - acc: 0.6121\n",
            "Epoch 9/50\n",
            "5130/5130 [==============================] - 1s 123us/sample - loss: 1.0711 - acc: 0.6199\n",
            "Epoch 10/50\n",
            "5130/5130 [==============================] - 1s 122us/sample - loss: 1.0482 - acc: 0.6251\n",
            "Epoch 11/50\n",
            "5130/5130 [==============================] - 1s 127us/sample - loss: 1.0207 - acc: 0.6378\n",
            "Epoch 12/50\n",
            "5130/5130 [==============================] - 1s 122us/sample - loss: 1.0032 - acc: 0.6437\n",
            "Epoch 13/50\n",
            "5130/5130 [==============================] - 1s 127us/sample - loss: 0.9797 - acc: 0.6515\n",
            "Epoch 14/50\n",
            "5130/5130 [==============================] - 1s 128us/sample - loss: 0.9676 - acc: 0.6600\n",
            "Epoch 15/50\n",
            "5130/5130 [==============================] - 1s 126us/sample - loss: 0.9460 - acc: 0.6680\n",
            "Epoch 16/50\n",
            "5130/5130 [==============================] - 1s 131us/sample - loss: 0.9364 - acc: 0.6704\n",
            "Epoch 17/50\n",
            "5130/5130 [==============================] - 1s 130us/sample - loss: 0.9221 - acc: 0.6713\n",
            "Epoch 18/50\n",
            "5130/5130 [==============================] - 1s 119us/sample - loss: 0.9158 - acc: 0.6780\n",
            "Epoch 19/50\n",
            "5130/5130 [==============================] - 1s 125us/sample - loss: 0.9041 - acc: 0.6834\n",
            "Epoch 20/50\n",
            "5130/5130 [==============================] - 1s 121us/sample - loss: 0.8969 - acc: 0.6858\n",
            "Epoch 21/50\n",
            "5130/5130 [==============================] - 1s 121us/sample - loss: 0.8882 - acc: 0.6924\n",
            "Epoch 22/50\n",
            "5130/5130 [==============================] - 1s 123us/sample - loss: 0.8733 - acc: 0.6942\n",
            "Epoch 23/50\n",
            "5130/5130 [==============================] - 1s 121us/sample - loss: 0.8707 - acc: 0.6930\n",
            "Epoch 24/50\n",
            "5130/5130 [==============================] - 1s 122us/sample - loss: 0.8592 - acc: 0.6979\n",
            "Epoch 25/50\n",
            "5130/5130 [==============================] - 1s 126us/sample - loss: 0.8475 - acc: 0.7043\n",
            "Epoch 26/50\n",
            "5130/5130 [==============================] - 1s 120us/sample - loss: 0.8492 - acc: 0.7041\n",
            "Epoch 27/50\n",
            "5130/5130 [==============================] - 1s 124us/sample - loss: 0.8320 - acc: 0.7029\n",
            "Epoch 28/50\n",
            "5130/5130 [==============================] - 1s 120us/sample - loss: 0.8057 - acc: 0.7214\n",
            "Epoch 29/50\n",
            "5130/5130 [==============================] - 1s 126us/sample - loss: 0.8203 - acc: 0.7201\n",
            "Epoch 30/50\n",
            "5130/5130 [==============================] - 1s 126us/sample - loss: 0.8065 - acc: 0.7162\n",
            "Epoch 31/50\n",
            "5130/5130 [==============================] - 1s 121us/sample - loss: 0.7959 - acc: 0.7265\n",
            "Epoch 32/50\n",
            "5130/5130 [==============================] - 1s 125us/sample - loss: 0.7994 - acc: 0.7148\n",
            "Epoch 33/50\n",
            "5130/5130 [==============================] - 1s 124us/sample - loss: 0.7911 - acc: 0.7255\n",
            "Epoch 34/50\n",
            "5130/5130 [==============================] - 1s 119us/sample - loss: 0.7972 - acc: 0.7173\n",
            "Epoch 35/50\n",
            "5130/5130 [==============================] - 1s 123us/sample - loss: 0.7805 - acc: 0.7279\n",
            "Epoch 36/50\n",
            "5130/5130 [==============================] - 1s 124us/sample - loss: 0.7700 - acc: 0.7248\n",
            "Epoch 37/50\n",
            "5130/5130 [==============================] - 1s 124us/sample - loss: 0.7725 - acc: 0.7273\n",
            "Epoch 38/50\n",
            "5130/5130 [==============================] - 1s 122us/sample - loss: 0.7726 - acc: 0.7279\n",
            "Epoch 39/50\n",
            "5130/5130 [==============================] - 1s 119us/sample - loss: 0.7630 - acc: 0.7370\n",
            "Epoch 40/50\n",
            "5130/5130 [==============================] - 1s 125us/sample - loss: 0.7600 - acc: 0.7331\n",
            "Epoch 41/50\n",
            "5130/5130 [==============================] - 1s 120us/sample - loss: 0.7514 - acc: 0.7355\n",
            "Epoch 42/50\n",
            "5130/5130 [==============================] - 1s 124us/sample - loss: 0.7495 - acc: 0.7386\n",
            "Epoch 43/50\n",
            "5130/5130 [==============================] - 1s 120us/sample - loss: 0.7440 - acc: 0.7298\n",
            "Epoch 44/50\n",
            "5130/5130 [==============================] - 1s 122us/sample - loss: 0.7300 - acc: 0.7417\n",
            "Epoch 45/50\n",
            "5130/5130 [==============================] - 1s 122us/sample - loss: 0.7314 - acc: 0.7462\n",
            "Epoch 46/50\n",
            "5130/5130 [==============================] - 1s 124us/sample - loss: 0.7315 - acc: 0.7437\n",
            "Epoch 47/50\n",
            "5130/5130 [==============================] - 1s 121us/sample - loss: 0.7176 - acc: 0.7468\n",
            "Epoch 48/50\n",
            "5130/5130 [==============================] - 1s 124us/sample - loss: 0.7267 - acc: 0.7433\n",
            "Epoch 49/50\n",
            "5130/5130 [==============================] - 1s 119us/sample - loss: 0.7043 - acc: 0.7550\n",
            "Epoch 50/50\n",
            "5130/5130 [==============================] - 1s 124us/sample - loss: 0.7140 - acc: 0.7515\n",
            "Epoch 1/50\n",
            "5130/5130 [==============================] - 1s 143us/sample - loss: 1.6959 - acc: 0.3971\n",
            "Epoch 2/50\n",
            "5130/5130 [==============================] - 1s 110us/sample - loss: 1.3955 - acc: 0.5072\n",
            "Epoch 3/50\n",
            "5130/5130 [==============================] - 1s 107us/sample - loss: 1.2913 - acc: 0.5388\n",
            "Epoch 4/50\n",
            "5130/5130 [==============================] - 1s 110us/sample - loss: 1.2377 - acc: 0.5558\n",
            "Epoch 5/50\n",
            "5130/5130 [==============================] - 1s 112us/sample - loss: 1.1868 - acc: 0.5770\n",
            "Epoch 6/50\n",
            "5130/5130 [==============================] - 1s 108us/sample - loss: 1.1616 - acc: 0.5871\n",
            "Epoch 7/50\n",
            "5130/5130 [==============================] - 1s 112us/sample - loss: 1.1164 - acc: 0.6002\n",
            "Epoch 8/50\n",
            "5130/5130 [==============================] - 1s 109us/sample - loss: 1.0829 - acc: 0.6232\n",
            "Epoch 9/50\n",
            "5130/5130 [==============================] - 1s 116us/sample - loss: 1.0662 - acc: 0.6205\n",
            "Epoch 10/50\n",
            "5130/5130 [==============================] - 1s 107us/sample - loss: 1.0460 - acc: 0.6296\n",
            "Epoch 11/50\n",
            "5130/5130 [==============================] - 1s 110us/sample - loss: 1.0155 - acc: 0.6442\n",
            "Epoch 12/50\n",
            "5130/5130 [==============================] - 1s 108us/sample - loss: 1.0067 - acc: 0.6464\n",
            "Epoch 13/50\n",
            "5130/5130 [==============================] - 1s 111us/sample - loss: 0.9907 - acc: 0.6487\n",
            "Epoch 14/50\n",
            "5130/5130 [==============================] - 1s 111us/sample - loss: 0.9745 - acc: 0.6596\n",
            "Epoch 15/50\n",
            "5130/5130 [==============================] - 1s 108us/sample - loss: 0.9647 - acc: 0.6616\n",
            "Epoch 16/50\n",
            "5130/5130 [==============================] - 1s 111us/sample - loss: 0.9536 - acc: 0.6571\n",
            "Epoch 17/50\n",
            "5130/5130 [==============================] - 1s 107us/sample - loss: 0.9426 - acc: 0.6754\n",
            "Epoch 18/50\n",
            "5130/5130 [==============================] - 1s 109us/sample - loss: 0.9222 - acc: 0.6795\n",
            "Epoch 19/50\n",
            "5130/5130 [==============================] - 1s 118us/sample - loss: 0.9188 - acc: 0.6741\n",
            "Epoch 20/50\n",
            "5130/5130 [==============================] - 1s 118us/sample - loss: 0.8952 - acc: 0.6906\n",
            "Epoch 21/50\n",
            "5130/5130 [==============================] - 1s 118us/sample - loss: 0.8993 - acc: 0.6842\n",
            "Epoch 22/50\n",
            "5130/5130 [==============================] - 1s 112us/sample - loss: 0.8902 - acc: 0.6932\n",
            "Epoch 23/50\n",
            "5130/5130 [==============================] - 1s 116us/sample - loss: 0.8726 - acc: 0.7029\n",
            "Epoch 24/50\n",
            "5130/5130 [==============================] - 1s 108us/sample - loss: 0.8839 - acc: 0.7019\n",
            "Epoch 25/50\n",
            "5130/5130 [==============================] - 1s 114us/sample - loss: 0.8516 - acc: 0.7068\n",
            "Epoch 26/50\n",
            "5130/5130 [==============================] - 1s 111us/sample - loss: 0.8448 - acc: 0.7058\n",
            "Epoch 27/50\n",
            "5130/5130 [==============================] - 1s 113us/sample - loss: 0.8453 - acc: 0.7064\n",
            "Epoch 28/50\n",
            "5130/5130 [==============================] - 1s 110us/sample - loss: 0.8369 - acc: 0.7088\n",
            "Epoch 29/50\n",
            "5130/5130 [==============================] - 1s 110us/sample - loss: 0.8360 - acc: 0.7105\n",
            "Epoch 30/50\n",
            "5130/5130 [==============================] - 1s 110us/sample - loss: 0.8202 - acc: 0.7191\n",
            "Epoch 31/50\n",
            "5130/5130 [==============================] - 1s 109us/sample - loss: 0.8181 - acc: 0.7279\n",
            "Epoch 32/50\n",
            "5130/5130 [==============================] - 1s 110us/sample - loss: 0.8039 - acc: 0.7288\n",
            "Epoch 33/50\n",
            "5130/5130 [==============================] - 1s 109us/sample - loss: 0.8071 - acc: 0.7228\n",
            "Epoch 34/50\n",
            "5130/5130 [==============================] - 1s 113us/sample - loss: 0.8013 - acc: 0.7248\n",
            "Epoch 35/50\n",
            "5130/5130 [==============================] - 1s 108us/sample - loss: 0.7889 - acc: 0.7331\n",
            "Epoch 36/50\n",
            "5130/5130 [==============================] - 1s 114us/sample - loss: 0.7760 - acc: 0.7310\n",
            "Epoch 37/50\n",
            "5130/5130 [==============================] - 1s 111us/sample - loss: 0.7823 - acc: 0.7302\n",
            "Epoch 38/50\n",
            "5130/5130 [==============================] - 1s 110us/sample - loss: 0.7835 - acc: 0.7296\n",
            "Epoch 39/50\n",
            "5130/5130 [==============================] - 1s 111us/sample - loss: 0.7739 - acc: 0.7366\n",
            "Epoch 40/50\n",
            "5130/5130 [==============================] - 1s 110us/sample - loss: 0.7720 - acc: 0.7359\n",
            "Epoch 41/50\n",
            "5130/5130 [==============================] - 1s 113us/sample - loss: 0.7540 - acc: 0.7351\n",
            "Epoch 42/50\n",
            "5130/5130 [==============================] - 1s 108us/sample - loss: 0.7667 - acc: 0.7339\n",
            "Epoch 43/50\n",
            "5130/5130 [==============================] - 1s 113us/sample - loss: 0.7603 - acc: 0.7380\n",
            "Epoch 44/50\n",
            "5130/5130 [==============================] - 1s 108us/sample - loss: 0.7544 - acc: 0.7409\n",
            "Epoch 45/50\n",
            "5130/5130 [==============================] - 1s 112us/sample - loss: 0.7409 - acc: 0.7394\n",
            "Epoch 46/50\n",
            "5130/5130 [==============================] - 1s 113us/sample - loss: 0.7468 - acc: 0.7435\n",
            "Epoch 47/50\n",
            "5130/5130 [==============================] - 1s 108us/sample - loss: 0.7427 - acc: 0.7427\n",
            "Epoch 48/50\n",
            "5130/5130 [==============================] - 1s 113us/sample - loss: 0.7345 - acc: 0.7536\n",
            "Epoch 49/50\n",
            "5130/5130 [==============================] - 1s 107us/sample - loss: 0.7391 - acc: 0.7472\n",
            "Epoch 50/50\n",
            "5130/5130 [==============================] - 1s 113us/sample - loss: 0.7311 - acc: 0.7454\n",
            "Epoch 1/50\n",
            "5130/5130 [==============================] - 1s 159us/sample - loss: 1.7566 - acc: 0.3750\n",
            "Epoch 2/50\n",
            "5130/5130 [==============================] - 1s 122us/sample - loss: 1.4271 - acc: 0.4992\n",
            "Epoch 3/50\n",
            "5130/5130 [==============================] - 1s 124us/sample - loss: 1.3140 - acc: 0.5380\n",
            "Epoch 4/50\n",
            "5130/5130 [==============================] - 1s 119us/sample - loss: 1.2540 - acc: 0.5540\n",
            "Epoch 5/50\n",
            "5130/5130 [==============================] - 1s 126us/sample - loss: 1.2012 - acc: 0.5733\n",
            "Epoch 6/50\n",
            "5130/5130 [==============================] - 1s 122us/sample - loss: 1.1722 - acc: 0.5776\n",
            "Epoch 7/50\n",
            "5130/5130 [==============================] - 1s 121us/sample - loss: 1.1276 - acc: 0.5998\n",
            "Epoch 8/50\n",
            "5130/5130 [==============================] - 1s 124us/sample - loss: 1.0995 - acc: 0.6113\n",
            "Epoch 9/50\n",
            "5130/5130 [==============================] - 1s 120us/sample - loss: 1.0648 - acc: 0.6173\n",
            "Epoch 10/50\n",
            "5130/5130 [==============================] - 1s 122us/sample - loss: 1.0461 - acc: 0.6279\n",
            "Epoch 11/50\n",
            "5130/5130 [==============================] - 1s 124us/sample - loss: 1.0248 - acc: 0.6368\n",
            "Epoch 12/50\n",
            "5130/5130 [==============================] - 1s 119us/sample - loss: 1.0219 - acc: 0.6405\n",
            "Epoch 13/50\n",
            "5130/5130 [==============================] - 1s 121us/sample - loss: 0.9880 - acc: 0.6493\n",
            "Epoch 14/50\n",
            "5130/5130 [==============================] - 1s 120us/sample - loss: 0.9790 - acc: 0.6565\n",
            "Epoch 15/50\n",
            "5130/5130 [==============================] - 1s 122us/sample - loss: 0.9667 - acc: 0.6606\n",
            "Epoch 16/50\n",
            "5130/5130 [==============================] - 1s 122us/sample - loss: 0.9439 - acc: 0.6585\n",
            "Epoch 17/50\n",
            "5130/5130 [==============================] - 1s 119us/sample - loss: 0.9273 - acc: 0.6704\n",
            "Epoch 18/50\n",
            "5130/5130 [==============================] - 1s 126us/sample - loss: 0.9301 - acc: 0.6721\n",
            "Epoch 19/50\n",
            "5130/5130 [==============================] - 1s 122us/sample - loss: 0.9135 - acc: 0.6832\n",
            "Epoch 20/50\n",
            "5130/5130 [==============================] - 1s 121us/sample - loss: 0.9132 - acc: 0.6768\n",
            "Epoch 21/50\n",
            "5130/5130 [==============================] - 1s 124us/sample - loss: 0.8967 - acc: 0.6840\n",
            "Epoch 22/50\n",
            "5130/5130 [==============================] - 1s 122us/sample - loss: 0.8913 - acc: 0.6838\n",
            "Epoch 23/50\n",
            "5130/5130 [==============================] - 1s 122us/sample - loss: 0.8641 - acc: 0.6967\n",
            "Epoch 24/50\n",
            "5130/5130 [==============================] - 1s 123us/sample - loss: 0.8710 - acc: 0.6982\n",
            "Epoch 25/50\n",
            "5130/5130 [==============================] - 1s 123us/sample - loss: 0.8653 - acc: 0.6943\n",
            "Epoch 26/50\n",
            "5130/5130 [==============================] - 1s 122us/sample - loss: 0.8501 - acc: 0.7018\n",
            "Epoch 27/50\n",
            "5130/5130 [==============================] - 1s 121us/sample - loss: 0.8506 - acc: 0.7037\n",
            "Epoch 28/50\n",
            "5130/5130 [==============================] - 1s 124us/sample - loss: 0.8364 - acc: 0.7086\n",
            "Epoch 29/50\n",
            "5130/5130 [==============================] - 1s 121us/sample - loss: 0.8188 - acc: 0.7140\n",
            "Epoch 30/50\n",
            "5130/5130 [==============================] - 1s 119us/sample - loss: 0.8246 - acc: 0.7097\n",
            "Epoch 31/50\n",
            "5130/5130 [==============================] - 1s 122us/sample - loss: 0.8217 - acc: 0.7111\n",
            "Epoch 32/50\n",
            "5130/5130 [==============================] - 1s 119us/sample - loss: 0.7956 - acc: 0.7189\n",
            "Epoch 33/50\n",
            "5130/5130 [==============================] - 1s 125us/sample - loss: 0.8037 - acc: 0.7148\n",
            "Epoch 34/50\n",
            "5130/5130 [==============================] - 1s 125us/sample - loss: 0.7910 - acc: 0.7183\n",
            "Epoch 35/50\n",
            "5130/5130 [==============================] - 1s 118us/sample - loss: 0.7926 - acc: 0.7281\n",
            "Epoch 36/50\n",
            "5130/5130 [==============================] - 1s 121us/sample - loss: 0.7924 - acc: 0.7164\n",
            "Epoch 37/50\n",
            "5130/5130 [==============================] - 1s 125us/sample - loss: 0.7753 - acc: 0.7288\n",
            "Epoch 38/50\n",
            "5130/5130 [==============================] - 1s 119us/sample - loss: 0.7796 - acc: 0.7273\n",
            "Epoch 39/50\n",
            "5130/5130 [==============================] - 1s 122us/sample - loss: 0.7614 - acc: 0.7372\n",
            "Epoch 40/50\n",
            "5130/5130 [==============================] - 1s 119us/sample - loss: 0.7557 - acc: 0.7349\n",
            "Epoch 41/50\n",
            "5130/5130 [==============================] - 1s 124us/sample - loss: 0.7471 - acc: 0.7386\n",
            "Epoch 42/50\n",
            "5130/5130 [==============================] - 1s 122us/sample - loss: 0.7357 - acc: 0.7427\n",
            "Epoch 43/50\n",
            "5130/5130 [==============================] - 1s 119us/sample - loss: 0.7621 - acc: 0.7267\n",
            "Epoch 44/50\n",
            "5130/5130 [==============================] - 1s 122us/sample - loss: 0.7293 - acc: 0.7480\n",
            "Epoch 45/50\n",
            "5130/5130 [==============================] - 1s 119us/sample - loss: 0.7375 - acc: 0.7372\n",
            "Epoch 46/50\n",
            "5130/5130 [==============================] - 1s 122us/sample - loss: 0.7343 - acc: 0.7396\n",
            "Epoch 47/50\n",
            "5130/5130 [==============================] - 1s 124us/sample - loss: 0.7225 - acc: 0.7404\n",
            "Epoch 48/50\n",
            "5130/5130 [==============================] - 1s 119us/sample - loss: 0.7229 - acc: 0.7483\n",
            "Epoch 49/50\n",
            "5130/5130 [==============================] - 1s 122us/sample - loss: 0.7162 - acc: 0.7540\n",
            "Epoch 50/50\n",
            "5130/5130 [==============================] - 1s 121us/sample - loss: 0.7093 - acc: 0.7528\n",
            "Epoch 1/50\n",
            "5130/5130 [==============================] - 1s 154us/sample - loss: 1.7025 - acc: 0.4045\n",
            "Epoch 2/50\n",
            "5130/5130 [==============================] - 1s 110us/sample - loss: 1.4036 - acc: 0.5058\n",
            "Epoch 3/50\n",
            "5130/5130 [==============================] - 1s 110us/sample - loss: 1.3137 - acc: 0.5355\n",
            "Epoch 4/50\n",
            "5130/5130 [==============================] - 1s 119us/sample - loss: 1.2404 - acc: 0.5552\n",
            "Epoch 5/50\n",
            "5130/5130 [==============================] - 1s 111us/sample - loss: 1.2062 - acc: 0.5678\n",
            "Epoch 6/50\n",
            "5130/5130 [==============================] - 1s 112us/sample - loss: 1.1549 - acc: 0.5832\n",
            "Epoch 7/50\n",
            "5130/5130 [==============================] - 1s 111us/sample - loss: 1.1378 - acc: 0.5920\n",
            "Epoch 8/50\n",
            "5130/5130 [==============================] - 1s 108us/sample - loss: 1.1092 - acc: 0.6023\n",
            "Epoch 9/50\n",
            "5130/5130 [==============================] - 1s 113us/sample - loss: 1.0752 - acc: 0.6191\n",
            "Epoch 10/50\n",
            "5130/5130 [==============================] - 1s 113us/sample - loss: 1.0549 - acc: 0.6248\n",
            "Epoch 11/50\n",
            "5130/5130 [==============================] - 1s 116us/sample - loss: 1.0410 - acc: 0.6255\n",
            "Epoch 12/50\n",
            "5130/5130 [==============================] - 1s 114us/sample - loss: 1.0198 - acc: 0.6398\n",
            "Epoch 13/50\n",
            "5130/5130 [==============================] - 1s 111us/sample - loss: 0.9943 - acc: 0.6536\n",
            "Epoch 14/50\n",
            "5130/5130 [==============================] - 1s 110us/sample - loss: 0.9825 - acc: 0.6538\n",
            "Epoch 15/50\n",
            "5130/5130 [==============================] - 1s 115us/sample - loss: 0.9649 - acc: 0.6606\n",
            "Epoch 16/50\n",
            "5130/5130 [==============================] - 1s 113us/sample - loss: 0.9559 - acc: 0.6616\n",
            "Epoch 17/50\n",
            "5130/5130 [==============================] - 1s 110us/sample - loss: 0.9528 - acc: 0.6618\n",
            "Epoch 18/50\n",
            "5130/5130 [==============================] - 1s 112us/sample - loss: 0.9449 - acc: 0.6674\n",
            "Epoch 19/50\n",
            "5130/5130 [==============================] - 1s 110us/sample - loss: 0.9378 - acc: 0.6712\n",
            "Epoch 20/50\n",
            "5130/5130 [==============================] - 1s 113us/sample - loss: 0.8987 - acc: 0.6823\n",
            "Epoch 21/50\n",
            "5130/5130 [==============================] - 1s 110us/sample - loss: 0.9058 - acc: 0.6809\n",
            "Epoch 22/50\n",
            "5130/5130 [==============================] - 1s 115us/sample - loss: 0.8893 - acc: 0.6840\n",
            "Epoch 23/50\n",
            "5130/5130 [==============================] - 1s 112us/sample - loss: 0.8965 - acc: 0.6819\n",
            "Epoch 24/50\n",
            "5130/5130 [==============================] - 1s 110us/sample - loss: 0.8723 - acc: 0.6967\n",
            "Epoch 25/50\n",
            "5130/5130 [==============================] - 1s 112us/sample - loss: 0.8703 - acc: 0.6977\n",
            "Epoch 26/50\n",
            "5130/5130 [==============================] - 1s 110us/sample - loss: 0.8576 - acc: 0.6969\n",
            "Epoch 27/50\n",
            "5130/5130 [==============================] - 1s 111us/sample - loss: 0.8548 - acc: 0.7010\n",
            "Epoch 28/50\n",
            "5130/5130 [==============================] - 1s 110us/sample - loss: 0.8307 - acc: 0.7140\n",
            "Epoch 29/50\n",
            "5130/5130 [==============================] - 1s 113us/sample - loss: 0.8399 - acc: 0.7088\n",
            "Epoch 30/50\n",
            "5130/5130 [==============================] - 1s 108us/sample - loss: 0.8436 - acc: 0.7019\n",
            "Epoch 31/50\n",
            "5130/5130 [==============================] - 1s 114us/sample - loss: 0.8207 - acc: 0.7099\n",
            "Epoch 32/50\n",
            "5130/5130 [==============================] - 1s 112us/sample - loss: 0.8189 - acc: 0.7173\n",
            "Epoch 33/50\n",
            "5130/5130 [==============================] - 1s 108us/sample - loss: 0.8116 - acc: 0.7117\n",
            "Epoch 34/50\n",
            "5130/5130 [==============================] - 1s 112us/sample - loss: 0.8208 - acc: 0.7168\n",
            "Epoch 35/50\n",
            "5130/5130 [==============================] - 1s 109us/sample - loss: 0.7963 - acc: 0.7308\n",
            "Epoch 36/50\n",
            "5130/5130 [==============================] - 1s 111us/sample - loss: 0.7938 - acc: 0.7240\n",
            "Epoch 37/50\n",
            "5130/5130 [==============================] - 1s 109us/sample - loss: 0.7849 - acc: 0.7226\n",
            "Epoch 38/50\n",
            "5130/5130 [==============================] - 1s 114us/sample - loss: 0.7829 - acc: 0.7250\n",
            "Epoch 39/50\n",
            "5130/5130 [==============================] - 1s 108us/sample - loss: 0.7853 - acc: 0.7246\n",
            "Epoch 40/50\n",
            "5130/5130 [==============================] - 1s 112us/sample - loss: 0.7785 - acc: 0.7287\n",
            "Epoch 41/50\n",
            "5130/5130 [==============================] - 1s 117us/sample - loss: 0.7641 - acc: 0.7318\n",
            "Epoch 42/50\n",
            "5130/5130 [==============================] - 1s 117us/sample - loss: 0.7743 - acc: 0.7240\n",
            "Epoch 43/50\n",
            "5130/5130 [==============================] - 1s 119us/sample - loss: 0.7639 - acc: 0.7347\n",
            "Epoch 44/50\n",
            "5130/5130 [==============================] - 1s 115us/sample - loss: 0.7589 - acc: 0.7372\n",
            "Epoch 45/50\n",
            "5130/5130 [==============================] - 1s 115us/sample - loss: 0.7602 - acc: 0.7400\n",
            "Epoch 46/50\n",
            "5130/5130 [==============================] - 1s 112us/sample - loss: 0.7533 - acc: 0.7304\n",
            "Epoch 47/50\n",
            "5130/5130 [==============================] - 1s 109us/sample - loss: 0.7548 - acc: 0.7384\n",
            "Epoch 48/50\n",
            "5130/5130 [==============================] - 1s 116us/sample - loss: 0.7422 - acc: 0.7425\n",
            "Epoch 49/50\n",
            "5130/5130 [==============================] - 1s 109us/sample - loss: 0.7294 - acc: 0.7480\n",
            "Epoch 50/50\n",
            "5130/5130 [==============================] - 1s 116us/sample - loss: 0.7333 - acc: 0.7439\n",
            "Epoch 1/50\n",
            "5130/5130 [==============================] - 1s 171us/sample - loss: 1.7529 - acc: 0.3789\n",
            "Epoch 2/50\n",
            "5130/5130 [==============================] - 1s 123us/sample - loss: 1.4081 - acc: 0.5111\n",
            "Epoch 3/50\n",
            "5130/5130 [==============================] - 1s 121us/sample - loss: 1.3123 - acc: 0.5398\n",
            "Epoch 4/50\n",
            "5130/5130 [==============================] - 1s 128us/sample - loss: 1.2469 - acc: 0.5593\n",
            "Epoch 5/50\n",
            "5130/5130 [==============================] - 1s 124us/sample - loss: 1.1943 - acc: 0.5764\n",
            "Epoch 6/50\n",
            "5130/5130 [==============================] - 1s 121us/sample - loss: 1.1510 - acc: 0.5951\n",
            "Epoch 7/50\n",
            "5130/5130 [==============================] - 1s 127us/sample - loss: 1.1262 - acc: 0.5984\n",
            "Epoch 8/50\n",
            "5130/5130 [==============================] - 1s 129us/sample - loss: 1.0891 - acc: 0.6099\n",
            "Epoch 9/50\n",
            "5130/5130 [==============================] - 1s 133us/sample - loss: 1.0681 - acc: 0.6226\n",
            "Epoch 10/50\n",
            "5130/5130 [==============================] - 1s 131us/sample - loss: 1.0396 - acc: 0.6220\n",
            "Epoch 11/50\n",
            "5130/5130 [==============================] - 1s 129us/sample - loss: 1.0155 - acc: 0.6485\n",
            "Epoch 12/50\n",
            "5130/5130 [==============================] - 1s 136us/sample - loss: 1.0004 - acc: 0.6478\n",
            "Epoch 13/50\n",
            "5130/5130 [==============================] - 1s 132us/sample - loss: 0.9829 - acc: 0.6524\n",
            "Epoch 14/50\n",
            "5130/5130 [==============================] - 1s 131us/sample - loss: 0.9668 - acc: 0.6595\n",
            "Epoch 15/50\n",
            "5130/5130 [==============================] - 1s 133us/sample - loss: 0.9572 - acc: 0.6622\n",
            "Epoch 16/50\n",
            "5130/5130 [==============================] - 1s 135us/sample - loss: 0.9479 - acc: 0.6663\n",
            "Epoch 17/50\n",
            "5130/5130 [==============================] - 1s 130us/sample - loss: 0.9347 - acc: 0.6673\n",
            "Epoch 18/50\n",
            "5130/5130 [==============================] - 1s 132us/sample - loss: 0.9159 - acc: 0.6807\n",
            "Epoch 19/50\n",
            "5130/5130 [==============================] - 1s 137us/sample - loss: 0.8974 - acc: 0.6854\n",
            "Epoch 20/50\n",
            "5130/5130 [==============================] - 1s 131us/sample - loss: 0.8989 - acc: 0.6825\n",
            "Epoch 21/50\n",
            "5130/5130 [==============================] - 1s 132us/sample - loss: 0.8777 - acc: 0.6910\n",
            "Epoch 22/50\n",
            "5130/5130 [==============================] - 1s 134us/sample - loss: 0.8697 - acc: 0.6934\n",
            "Epoch 23/50\n",
            "5130/5130 [==============================] - 1s 126us/sample - loss: 0.8623 - acc: 0.6981\n",
            "Epoch 24/50\n",
            "5130/5130 [==============================] - 1s 123us/sample - loss: 0.8534 - acc: 0.7041\n",
            "Epoch 25/50\n",
            "5130/5130 [==============================] - 1s 123us/sample - loss: 0.8517 - acc: 0.7041\n",
            "Epoch 26/50\n",
            "5130/5130 [==============================] - 1s 121us/sample - loss: 0.8234 - acc: 0.7131\n",
            "Epoch 27/50\n",
            "5130/5130 [==============================] - 1s 126us/sample - loss: 0.8265 - acc: 0.7103\n",
            "Epoch 28/50\n",
            "5130/5130 [==============================] - 1s 122us/sample - loss: 0.8248 - acc: 0.7144\n",
            "Epoch 29/50\n",
            "5130/5130 [==============================] - 1s 125us/sample - loss: 0.8125 - acc: 0.7117\n",
            "Epoch 30/50\n",
            "5130/5130 [==============================] - 1s 123us/sample - loss: 0.8194 - acc: 0.7072\n",
            "Epoch 31/50\n",
            "5130/5130 [==============================] - 1s 120us/sample - loss: 0.8030 - acc: 0.7138\n",
            "Epoch 32/50\n",
            "5130/5130 [==============================] - 1s 123us/sample - loss: 0.8083 - acc: 0.7216\n",
            "Epoch 33/50\n",
            "5130/5130 [==============================] - 1s 125us/sample - loss: 0.7958 - acc: 0.7269\n",
            "Epoch 34/50\n",
            "5130/5130 [==============================] - 1s 122us/sample - loss: 0.7879 - acc: 0.7220\n",
            "Epoch 35/50\n",
            "5130/5130 [==============================] - 1s 125us/sample - loss: 0.7843 - acc: 0.7283\n",
            "Epoch 36/50\n",
            "5130/5130 [==============================] - 1s 121us/sample - loss: 0.7735 - acc: 0.7331\n",
            "Epoch 37/50\n",
            "5130/5130 [==============================] - 1s 124us/sample - loss: 0.7630 - acc: 0.7312\n",
            "Epoch 38/50\n",
            "5130/5130 [==============================] - 1s 125us/sample - loss: 0.7648 - acc: 0.7279\n",
            "Epoch 39/50\n",
            "5130/5130 [==============================] - 1s 121us/sample - loss: 0.7671 - acc: 0.7288\n",
            "Epoch 40/50\n",
            "5130/5130 [==============================] - 1s 125us/sample - loss: 0.7594 - acc: 0.7329\n",
            "Epoch 41/50\n",
            "5130/5130 [==============================] - 1s 123us/sample - loss: 0.7484 - acc: 0.7429\n",
            "Epoch 42/50\n",
            "5130/5130 [==============================] - 1s 125us/sample - loss: 0.7529 - acc: 0.7316\n",
            "Epoch 43/50\n",
            "5130/5130 [==============================] - 1s 124us/sample - loss: 0.7497 - acc: 0.7384\n",
            "Epoch 44/50\n",
            "5130/5130 [==============================] - 1s 121us/sample - loss: 0.7368 - acc: 0.7366\n",
            "Epoch 45/50\n",
            "5130/5130 [==============================] - 1s 128us/sample - loss: 0.7316 - acc: 0.7427\n",
            "Epoch 46/50\n",
            "5130/5130 [==============================] - 1s 126us/sample - loss: 0.7253 - acc: 0.7458\n",
            "Epoch 47/50\n",
            "5130/5130 [==============================] - 1s 123us/sample - loss: 0.7179 - acc: 0.7480\n",
            "Epoch 48/50\n",
            "5130/5130 [==============================] - 1s 124us/sample - loss: 0.7239 - acc: 0.7470\n",
            "Epoch 49/50\n",
            "5130/5130 [==============================] - 1s 125us/sample - loss: 0.7070 - acc: 0.7552\n",
            "Epoch 50/50\n",
            "5130/5130 [==============================] - 1s 123us/sample - loss: 0.7193 - acc: 0.7517\n",
            "Epoch 1/50\n",
            "5130/5130 [==============================] - 1s 101us/sample - loss: 1.8087 - acc: 0.3686\n",
            "Epoch 2/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 1.4727 - acc: 0.4926\n",
            "Epoch 3/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 1.3691 - acc: 0.5203\n",
            "Epoch 4/50\n",
            "5130/5130 [==============================] - 0s 60us/sample - loss: 1.2988 - acc: 0.5417\n",
            "Epoch 5/50\n",
            "5130/5130 [==============================] - 0s 55us/sample - loss: 1.2480 - acc: 0.5563\n",
            "Epoch 6/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 1.2149 - acc: 0.5647\n",
            "Epoch 7/50\n",
            "5130/5130 [==============================] - 0s 58us/sample - loss: 1.1914 - acc: 0.5754\n",
            "Epoch 8/50\n",
            "5130/5130 [==============================] - 0s 59us/sample - loss: 1.1548 - acc: 0.5953\n",
            "Epoch 9/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 1.1274 - acc: 0.5986\n",
            "Epoch 10/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 1.1104 - acc: 0.6117\n",
            "Epoch 11/50\n",
            "5130/5130 [==============================] - 0s 58us/sample - loss: 1.0838 - acc: 0.6135\n",
            "Epoch 12/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 1.0725 - acc: 0.6172\n",
            "Epoch 13/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 1.0496 - acc: 0.6312\n",
            "Epoch 14/50\n",
            "5130/5130 [==============================] - 0s 59us/sample - loss: 1.0342 - acc: 0.6423\n",
            "Epoch 15/50\n",
            "5130/5130 [==============================] - 0s 58us/sample - loss: 1.0167 - acc: 0.6425\n",
            "Epoch 16/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 1.0084 - acc: 0.6468\n",
            "Epoch 17/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 1.0004 - acc: 0.6466\n",
            "Epoch 18/50\n",
            "5130/5130 [==============================] - 0s 58us/sample - loss: 0.9779 - acc: 0.6550\n",
            "Epoch 19/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 0.9811 - acc: 0.6456\n",
            "Epoch 20/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 0.9646 - acc: 0.6610\n",
            "Epoch 21/50\n",
            "5130/5130 [==============================] - 0s 58us/sample - loss: 0.9502 - acc: 0.6667\n",
            "Epoch 22/50\n",
            "5130/5130 [==============================] - 0s 55us/sample - loss: 0.9492 - acc: 0.6608\n",
            "Epoch 23/50\n",
            "5130/5130 [==============================] - 0s 55us/sample - loss: 0.9259 - acc: 0.6706\n",
            "Epoch 24/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 0.9301 - acc: 0.6704\n",
            "Epoch 25/50\n",
            "5130/5130 [==============================] - 0s 58us/sample - loss: 0.9277 - acc: 0.6805\n",
            "Epoch 26/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 0.9099 - acc: 0.6848\n",
            "Epoch 27/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 0.8955 - acc: 0.6881\n",
            "Epoch 28/50\n",
            "5130/5130 [==============================] - 0s 58us/sample - loss: 0.8912 - acc: 0.6920\n",
            "Epoch 29/50\n",
            "5130/5130 [==============================] - 0s 55us/sample - loss: 0.8824 - acc: 0.6969\n",
            "Epoch 30/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 0.8811 - acc: 0.6961\n",
            "Epoch 31/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 0.8696 - acc: 0.7039\n",
            "Epoch 32/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 0.8590 - acc: 0.6996\n",
            "Epoch 33/50\n",
            "5130/5130 [==============================] - 0s 54us/sample - loss: 0.8525 - acc: 0.7062\n",
            "Epoch 34/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 0.8648 - acc: 0.7004\n",
            "Epoch 35/50\n",
            "5130/5130 [==============================] - 0s 59us/sample - loss: 0.8537 - acc: 0.6973\n",
            "Epoch 36/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 0.8508 - acc: 0.7041\n",
            "Epoch 37/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 0.8513 - acc: 0.7053\n",
            "Epoch 38/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 0.8264 - acc: 0.7127\n",
            "Epoch 39/50\n",
            "5130/5130 [==============================] - 0s 59us/sample - loss: 0.8317 - acc: 0.7117\n",
            "Epoch 40/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 0.8222 - acc: 0.7158\n",
            "Epoch 41/50\n",
            "5130/5130 [==============================] - 0s 55us/sample - loss: 0.8252 - acc: 0.7175\n",
            "Epoch 42/50\n",
            "5130/5130 [==============================] - 0s 59us/sample - loss: 0.8175 - acc: 0.7125\n",
            "Epoch 43/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 0.8023 - acc: 0.7259\n",
            "Epoch 44/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 0.7910 - acc: 0.7250\n",
            "Epoch 45/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 0.8131 - acc: 0.7162\n",
            "Epoch 46/50\n",
            "5130/5130 [==============================] - 0s 58us/sample - loss: 0.7809 - acc: 0.7290\n",
            "Epoch 47/50\n",
            "5130/5130 [==============================] - 0s 58us/sample - loss: 0.7923 - acc: 0.7226\n",
            "Epoch 48/50\n",
            "5130/5130 [==============================] - 0s 58us/sample - loss: 0.7831 - acc: 0.7288\n",
            "Epoch 49/50\n",
            "5130/5130 [==============================] - 0s 60us/sample - loss: 0.7806 - acc: 0.7279\n",
            "Epoch 50/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 0.7857 - acc: 0.7253\n",
            "Epoch 1/50\n",
            "5130/5130 [==============================] - 1s 111us/sample - loss: 1.8325 - acc: 0.3558\n",
            "Epoch 2/50\n",
            "5130/5130 [==============================] - 0s 61us/sample - loss: 1.4910 - acc: 0.4838\n",
            "Epoch 3/50\n",
            "5130/5130 [==============================] - 0s 61us/sample - loss: 1.3682 - acc: 0.5121\n",
            "Epoch 4/50\n",
            "5130/5130 [==============================] - 0s 65us/sample - loss: 1.3008 - acc: 0.5441\n",
            "Epoch 5/50\n",
            "5130/5130 [==============================] - 0s 61us/sample - loss: 1.2528 - acc: 0.5546\n",
            "Epoch 6/50\n",
            "5130/5130 [==============================] - 0s 63us/sample - loss: 1.2156 - acc: 0.5674\n",
            "Epoch 7/50\n",
            "5130/5130 [==============================] - 0s 66us/sample - loss: 1.1866 - acc: 0.5805\n",
            "Epoch 8/50\n",
            "5130/5130 [==============================] - 0s 63us/sample - loss: 1.1592 - acc: 0.5887\n",
            "Epoch 9/50\n",
            "5130/5130 [==============================] - 0s 61us/sample - loss: 1.1257 - acc: 0.6004\n",
            "Epoch 10/50\n",
            "5130/5130 [==============================] - 0s 65us/sample - loss: 1.1034 - acc: 0.6016\n",
            "Epoch 11/50\n",
            "5130/5130 [==============================] - 0s 62us/sample - loss: 1.0932 - acc: 0.6222\n",
            "Epoch 12/50\n",
            "5130/5130 [==============================] - 0s 60us/sample - loss: 1.0667 - acc: 0.6207\n",
            "Epoch 13/50\n",
            "5130/5130 [==============================] - 0s 66us/sample - loss: 1.0531 - acc: 0.6201\n",
            "Epoch 14/50\n",
            "5130/5130 [==============================] - 0s 64us/sample - loss: 1.0444 - acc: 0.6322\n",
            "Epoch 15/50\n",
            "5130/5130 [==============================] - 0s 62us/sample - loss: 1.0149 - acc: 0.6296\n",
            "Epoch 16/50\n",
            "5130/5130 [==============================] - 0s 63us/sample - loss: 1.0089 - acc: 0.6415\n",
            "Epoch 17/50\n",
            "5130/5130 [==============================] - 0s 61us/sample - loss: 0.9999 - acc: 0.6454\n",
            "Epoch 18/50\n",
            "5130/5130 [==============================] - 0s 62us/sample - loss: 0.9712 - acc: 0.6585\n",
            "Epoch 19/50\n",
            "5130/5130 [==============================] - 0s 64us/sample - loss: 0.9745 - acc: 0.6575\n",
            "Epoch 20/50\n",
            "5130/5130 [==============================] - 0s 61us/sample - loss: 0.9639 - acc: 0.6606\n",
            "Epoch 21/50\n",
            "5130/5130 [==============================] - 0s 68us/sample - loss: 0.9544 - acc: 0.6583\n",
            "Epoch 22/50\n",
            "5130/5130 [==============================] - 0s 70us/sample - loss: 0.9453 - acc: 0.6698\n",
            "Epoch 23/50\n",
            "5130/5130 [==============================] - 0s 67us/sample - loss: 0.9263 - acc: 0.6770\n",
            "Epoch 24/50\n",
            "5130/5130 [==============================] - 0s 67us/sample - loss: 0.9298 - acc: 0.6754\n",
            "Epoch 25/50\n",
            "5130/5130 [==============================] - 0s 70us/sample - loss: 0.9103 - acc: 0.6774\n",
            "Epoch 26/50\n",
            "5130/5130 [==============================] - 0s 66us/sample - loss: 0.9094 - acc: 0.6819\n",
            "Epoch 27/50\n",
            "5130/5130 [==============================] - 0s 66us/sample - loss: 0.9012 - acc: 0.6893\n",
            "Epoch 28/50\n",
            "5130/5130 [==============================] - 0s 69us/sample - loss: 0.9021 - acc: 0.6823\n",
            "Epoch 29/50\n",
            "5130/5130 [==============================] - 0s 69us/sample - loss: 0.8827 - acc: 0.6951\n",
            "Epoch 30/50\n",
            "5130/5130 [==============================] - 0s 67us/sample - loss: 0.8788 - acc: 0.6912\n",
            "Epoch 31/50\n",
            "5130/5130 [==============================] - 0s 69us/sample - loss: 0.8622 - acc: 0.6996\n",
            "Epoch 32/50\n",
            "5130/5130 [==============================] - 0s 66us/sample - loss: 0.8695 - acc: 0.7043\n",
            "Epoch 33/50\n",
            "5130/5130 [==============================] - 0s 67us/sample - loss: 0.8773 - acc: 0.6904\n",
            "Epoch 34/50\n",
            "5130/5130 [==============================] - 0s 68us/sample - loss: 0.8515 - acc: 0.7018\n",
            "Epoch 35/50\n",
            "5130/5130 [==============================] - 0s 67us/sample - loss: 0.8335 - acc: 0.7105\n",
            "Epoch 36/50\n",
            "5130/5130 [==============================] - 0s 66us/sample - loss: 0.8523 - acc: 0.7010\n",
            "Epoch 37/50\n",
            "5130/5130 [==============================] - 0s 65us/sample - loss: 0.8388 - acc: 0.7175\n",
            "Epoch 38/50\n",
            "5130/5130 [==============================] - 0s 61us/sample - loss: 0.8356 - acc: 0.7084\n",
            "Epoch 39/50\n",
            "5130/5130 [==============================] - 0s 62us/sample - loss: 0.8242 - acc: 0.7127\n",
            "Epoch 40/50\n",
            "5130/5130 [==============================] - 0s 65us/sample - loss: 0.8200 - acc: 0.7162\n",
            "Epoch 41/50\n",
            "5130/5130 [==============================] - 0s 61us/sample - loss: 0.8058 - acc: 0.7201\n",
            "Epoch 42/50\n",
            "5130/5130 [==============================] - 0s 61us/sample - loss: 0.8159 - acc: 0.7168\n",
            "Epoch 43/50\n",
            "5130/5130 [==============================] - 0s 64us/sample - loss: 0.8069 - acc: 0.7185\n",
            "Epoch 44/50\n",
            "5130/5130 [==============================] - 0s 60us/sample - loss: 0.8026 - acc: 0.7207\n",
            "Epoch 45/50\n",
            "5130/5130 [==============================] - 0s 62us/sample - loss: 0.7977 - acc: 0.7187\n",
            "Epoch 46/50\n",
            "5130/5130 [==============================] - 0s 67us/sample - loss: 0.8043 - acc: 0.7212\n",
            "Epoch 47/50\n",
            "5130/5130 [==============================] - 0s 64us/sample - loss: 0.7943 - acc: 0.7222\n",
            "Epoch 48/50\n",
            "5130/5130 [==============================] - 0s 61us/sample - loss: 0.7898 - acc: 0.7312\n",
            "Epoch 49/50\n",
            "5130/5130 [==============================] - 0s 62us/sample - loss: 0.7915 - acc: 0.7212\n",
            "Epoch 50/50\n",
            "5130/5130 [==============================] - 0s 62us/sample - loss: 0.7853 - acc: 0.7283\n",
            "Epoch 1/50\n",
            "5130/5130 [==============================] - 1s 108us/sample - loss: 1.8061 - acc: 0.3614\n",
            "Epoch 2/50\n",
            "5130/5130 [==============================] - 0s 58us/sample - loss: 1.4856 - acc: 0.4745\n",
            "Epoch 3/50\n",
            "5130/5130 [==============================] - 0s 58us/sample - loss: 1.3617 - acc: 0.5234\n",
            "Epoch 4/50\n",
            "5130/5130 [==============================] - 0s 59us/sample - loss: 1.2952 - acc: 0.5450\n",
            "Epoch 5/50\n",
            "5130/5130 [==============================] - 0s 59us/sample - loss: 1.2453 - acc: 0.5614\n",
            "Epoch 6/50\n",
            "5130/5130 [==============================] - 0s 60us/sample - loss: 1.2088 - acc: 0.5768\n",
            "Epoch 7/50\n",
            "5130/5130 [==============================] - 0s 58us/sample - loss: 1.1870 - acc: 0.5805\n",
            "Epoch 8/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 1.1583 - acc: 0.5846\n",
            "Epoch 9/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 1.1360 - acc: 0.5977\n",
            "Epoch 10/50\n",
            "5130/5130 [==============================] - 0s 59us/sample - loss: 1.1105 - acc: 0.6047\n",
            "Epoch 11/50\n",
            "5130/5130 [==============================] - 0s 58us/sample - loss: 1.0860 - acc: 0.6158\n",
            "Epoch 12/50\n",
            "5130/5130 [==============================] - 0s 58us/sample - loss: 1.0797 - acc: 0.6154\n",
            "Epoch 13/50\n",
            "5130/5130 [==============================] - 0s 61us/sample - loss: 1.0519 - acc: 0.6222\n",
            "Epoch 14/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 1.0408 - acc: 0.6298\n",
            "Epoch 15/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 1.0247 - acc: 0.6378\n",
            "Epoch 16/50\n",
            "5130/5130 [==============================] - 0s 58us/sample - loss: 1.0143 - acc: 0.6439\n",
            "Epoch 17/50\n",
            "5130/5130 [==============================] - 0s 58us/sample - loss: 0.9975 - acc: 0.6515\n",
            "Epoch 18/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 0.9860 - acc: 0.6550\n",
            "Epoch 19/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 0.9910 - acc: 0.6487\n",
            "Epoch 20/50\n",
            "5130/5130 [==============================] - 0s 58us/sample - loss: 0.9650 - acc: 0.6614\n",
            "Epoch 21/50\n",
            "5130/5130 [==============================] - 0s 58us/sample - loss: 0.9641 - acc: 0.6669\n",
            "Epoch 22/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 0.9482 - acc: 0.6608\n",
            "Epoch 23/50\n",
            "5130/5130 [==============================] - 0s 59us/sample - loss: 0.9375 - acc: 0.6682\n",
            "Epoch 24/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 0.9355 - acc: 0.6659\n",
            "Epoch 25/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 0.9269 - acc: 0.6739\n",
            "Epoch 26/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 0.9070 - acc: 0.6817\n",
            "Epoch 27/50\n",
            "5130/5130 [==============================] - 0s 60us/sample - loss: 0.9123 - acc: 0.6801\n",
            "Epoch 28/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 0.9055 - acc: 0.6834\n",
            "Epoch 29/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 0.8753 - acc: 0.6940\n",
            "Epoch 30/50\n",
            "5130/5130 [==============================] - 0s 58us/sample - loss: 0.8726 - acc: 0.6994\n",
            "Epoch 31/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 0.8737 - acc: 0.6904\n",
            "Epoch 32/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 0.8721 - acc: 0.6914\n",
            "Epoch 33/50\n",
            "5130/5130 [==============================] - 0s 58us/sample - loss: 0.8640 - acc: 0.6988\n",
            "Epoch 34/50\n",
            "5130/5130 [==============================] - 0s 58us/sample - loss: 0.8488 - acc: 0.6973\n",
            "Epoch 35/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 0.8666 - acc: 0.6967\n",
            "Epoch 36/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 0.8484 - acc: 0.6959\n",
            "Epoch 37/50\n",
            "5130/5130 [==============================] - 0s 58us/sample - loss: 0.8425 - acc: 0.7018\n",
            "Epoch 38/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 0.8325 - acc: 0.7103\n",
            "Epoch 39/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 0.8286 - acc: 0.7170\n",
            "Epoch 40/50\n",
            "5130/5130 [==============================] - 0s 59us/sample - loss: 0.8381 - acc: 0.7103\n",
            "Epoch 41/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 0.8199 - acc: 0.7158\n",
            "Epoch 42/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 0.8191 - acc: 0.7195\n",
            "Epoch 43/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 0.8211 - acc: 0.7144\n",
            "Epoch 44/50\n",
            "5130/5130 [==============================] - 0s 59us/sample - loss: 0.8082 - acc: 0.7203\n",
            "Epoch 45/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 0.8092 - acc: 0.7166\n",
            "Epoch 46/50\n",
            "5130/5130 [==============================] - 0s 58us/sample - loss: 0.8000 - acc: 0.7216\n",
            "Epoch 47/50\n",
            "5130/5130 [==============================] - 0s 60us/sample - loss: 0.7912 - acc: 0.7214\n",
            "Epoch 48/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 0.7949 - acc: 0.7172\n",
            "Epoch 49/50\n",
            "5130/5130 [==============================] - 0s 59us/sample - loss: 0.7794 - acc: 0.7275\n",
            "Epoch 50/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 0.7811 - acc: 0.7298\n",
            "Epoch 1/50\n",
            "5130/5130 [==============================] - 1s 112us/sample - loss: 1.8718 - acc: 0.3366\n",
            "Epoch 2/50\n",
            "5130/5130 [==============================] - 0s 61us/sample - loss: 1.5051 - acc: 0.4741\n",
            "Epoch 3/50\n",
            "5130/5130 [==============================] - 0s 61us/sample - loss: 1.3823 - acc: 0.5211\n",
            "Epoch 4/50\n",
            "5130/5130 [==============================] - 0s 66us/sample - loss: 1.3077 - acc: 0.5368\n",
            "Epoch 5/50\n",
            "5130/5130 [==============================] - 0s 63us/sample - loss: 1.2562 - acc: 0.5559\n",
            "Epoch 6/50\n",
            "5130/5130 [==============================] - 0s 59us/sample - loss: 1.2208 - acc: 0.5725\n",
            "Epoch 7/50\n",
            "5130/5130 [==============================] - 0s 65us/sample - loss: 1.1877 - acc: 0.5756\n",
            "Epoch 8/50\n",
            "5130/5130 [==============================] - 0s 62us/sample - loss: 1.1486 - acc: 0.5924\n",
            "Epoch 9/50\n",
            "5130/5130 [==============================] - 0s 63us/sample - loss: 1.1345 - acc: 0.5951\n",
            "Epoch 10/50\n",
            "5130/5130 [==============================] - 0s 65us/sample - loss: 1.1116 - acc: 0.6051\n",
            "Epoch 11/50\n",
            "5130/5130 [==============================] - 0s 62us/sample - loss: 1.0811 - acc: 0.6179\n",
            "Epoch 12/50\n",
            "5130/5130 [==============================] - 0s 63us/sample - loss: 1.0643 - acc: 0.6185\n",
            "Epoch 13/50\n",
            "5130/5130 [==============================] - 0s 65us/sample - loss: 1.0450 - acc: 0.6248\n",
            "Epoch 14/50\n",
            "5130/5130 [==============================] - 0s 64us/sample - loss: 1.0341 - acc: 0.6271\n",
            "Epoch 15/50\n",
            "5130/5130 [==============================] - 0s 60us/sample - loss: 1.0155 - acc: 0.6370\n",
            "Epoch 16/50\n",
            "5130/5130 [==============================] - 0s 63us/sample - loss: 1.0028 - acc: 0.6433\n",
            "Epoch 17/50\n",
            "5130/5130 [==============================] - 0s 64us/sample - loss: 0.9884 - acc: 0.6487\n",
            "Epoch 18/50\n",
            "5130/5130 [==============================] - 0s 61us/sample - loss: 0.9820 - acc: 0.6552\n",
            "Epoch 19/50\n",
            "5130/5130 [==============================] - 0s 61us/sample - loss: 0.9694 - acc: 0.6616\n",
            "Epoch 20/50\n",
            "5130/5130 [==============================] - 0s 67us/sample - loss: 0.9575 - acc: 0.6678\n",
            "Epoch 21/50\n",
            "5130/5130 [==============================] - 0s 62us/sample - loss: 0.9438 - acc: 0.6645\n",
            "Epoch 22/50\n",
            "5130/5130 [==============================] - 0s 61us/sample - loss: 0.9381 - acc: 0.6749\n",
            "Epoch 23/50\n",
            "5130/5130 [==============================] - 0s 64us/sample - loss: 0.9266 - acc: 0.6712\n",
            "Epoch 24/50\n",
            "5130/5130 [==============================] - 0s 65us/sample - loss: 0.9208 - acc: 0.6805\n",
            "Epoch 25/50\n",
            "5130/5130 [==============================] - 0s 63us/sample - loss: 0.9145 - acc: 0.6778\n",
            "Epoch 26/50\n",
            "5130/5130 [==============================] - 0s 64us/sample - loss: 0.9057 - acc: 0.6817\n",
            "Epoch 27/50\n",
            "5130/5130 [==============================] - 0s 63us/sample - loss: 0.8960 - acc: 0.6830\n",
            "Epoch 28/50\n",
            "5130/5130 [==============================] - 0s 62us/sample - loss: 0.8958 - acc: 0.6936\n",
            "Epoch 29/50\n",
            "5130/5130 [==============================] - 0s 65us/sample - loss: 0.8792 - acc: 0.6901\n",
            "Epoch 30/50\n",
            "5130/5130 [==============================] - 0s 63us/sample - loss: 0.8744 - acc: 0.6940\n",
            "Epoch 31/50\n",
            "5130/5130 [==============================] - 0s 62us/sample - loss: 0.8626 - acc: 0.6940\n",
            "Epoch 32/50\n",
            "5130/5130 [==============================] - 0s 64us/sample - loss: 0.8743 - acc: 0.6932\n",
            "Epoch 33/50\n",
            "5130/5130 [==============================] - 0s 61us/sample - loss: 0.8647 - acc: 0.6982\n",
            "Epoch 34/50\n",
            "5130/5130 [==============================] - 0s 61us/sample - loss: 0.8540 - acc: 0.7055\n",
            "Epoch 35/50\n",
            "5130/5130 [==============================] - 0s 65us/sample - loss: 0.8378 - acc: 0.7094\n",
            "Epoch 36/50\n",
            "5130/5130 [==============================] - 0s 61us/sample - loss: 0.8325 - acc: 0.7133\n",
            "Epoch 37/50\n",
            "5130/5130 [==============================] - 0s 62us/sample - loss: 0.8239 - acc: 0.7082\n",
            "Epoch 38/50\n",
            "5130/5130 [==============================] - 0s 65us/sample - loss: 0.8152 - acc: 0.7168\n",
            "Epoch 39/50\n",
            "5130/5130 [==============================] - 0s 64us/sample - loss: 0.8207 - acc: 0.7115\n",
            "Epoch 40/50\n",
            "5130/5130 [==============================] - 0s 63us/sample - loss: 0.8131 - acc: 0.7109\n",
            "Epoch 41/50\n",
            "5130/5130 [==============================] - 0s 66us/sample - loss: 0.8145 - acc: 0.7152\n",
            "Epoch 42/50\n",
            "5130/5130 [==============================] - 0s 62us/sample - loss: 0.8147 - acc: 0.7060\n",
            "Epoch 43/50\n",
            "5130/5130 [==============================] - 0s 62us/sample - loss: 0.7891 - acc: 0.7300\n",
            "Epoch 44/50\n",
            "5130/5130 [==============================] - 0s 61us/sample - loss: 0.7869 - acc: 0.7257\n",
            "Epoch 45/50\n",
            "5130/5130 [==============================] - 0s 65us/sample - loss: 0.7857 - acc: 0.7279\n",
            "Epoch 46/50\n",
            "5130/5130 [==============================] - 0s 62us/sample - loss: 0.7833 - acc: 0.7288\n",
            "Epoch 47/50\n",
            "5130/5130 [==============================] - 0s 60us/sample - loss: 0.7780 - acc: 0.7312\n",
            "Epoch 48/50\n",
            "5130/5130 [==============================] - 0s 65us/sample - loss: 0.7661 - acc: 0.7388\n",
            "Epoch 49/50\n",
            "5130/5130 [==============================] - 0s 61us/sample - loss: 0.7654 - acc: 0.7294\n",
            "Epoch 50/50\n",
            "5130/5130 [==============================] - 0s 61us/sample - loss: 0.7656 - acc: 0.7314\n",
            "Epoch 1/50\n",
            "5130/5130 [==============================] - 1s 109us/sample - loss: 1.8024 - acc: 0.3643\n",
            "Epoch 2/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 1.4784 - acc: 0.4813\n",
            "Epoch 3/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 1.3639 - acc: 0.5222\n",
            "Epoch 4/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 1.3009 - acc: 0.5372\n",
            "Epoch 5/50\n",
            "5130/5130 [==============================] - 0s 59us/sample - loss: 1.2525 - acc: 0.5559\n",
            "Epoch 6/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 1.2091 - acc: 0.5645\n",
            "Epoch 7/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 1.1888 - acc: 0.5758\n",
            "Epoch 8/50\n",
            "5130/5130 [==============================] - 0s 60us/sample - loss: 1.1581 - acc: 0.5899\n",
            "Epoch 9/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 1.1228 - acc: 0.6033\n",
            "Epoch 10/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 1.1087 - acc: 0.6086\n",
            "Epoch 11/50\n",
            "5130/5130 [==============================] - 0s 58us/sample - loss: 1.0947 - acc: 0.6090\n",
            "Epoch 12/50\n",
            "5130/5130 [==============================] - 0s 59us/sample - loss: 1.0750 - acc: 0.6189\n",
            "Epoch 13/50\n",
            "5130/5130 [==============================] - 0s 58us/sample - loss: 1.0535 - acc: 0.6326\n",
            "Epoch 14/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 1.0363 - acc: 0.6337\n",
            "Epoch 15/50\n",
            "5130/5130 [==============================] - 0s 59us/sample - loss: 1.0202 - acc: 0.6333\n",
            "Epoch 16/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 1.0128 - acc: 0.6407\n",
            "Epoch 17/50\n",
            "5130/5130 [==============================] - 0s 58us/sample - loss: 0.9968 - acc: 0.6433\n",
            "Epoch 18/50\n",
            "5130/5130 [==============================] - 0s 59us/sample - loss: 0.9876 - acc: 0.6552\n",
            "Epoch 19/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 0.9789 - acc: 0.6618\n",
            "Epoch 20/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 0.9698 - acc: 0.6554\n",
            "Epoch 21/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 0.9649 - acc: 0.6577\n",
            "Epoch 22/50\n",
            "5130/5130 [==============================] - 0s 59us/sample - loss: 0.9452 - acc: 0.6649\n",
            "Epoch 23/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 0.9438 - acc: 0.6641\n",
            "Epoch 24/50\n",
            "5130/5130 [==============================] - 0s 58us/sample - loss: 0.9289 - acc: 0.6680\n",
            "Epoch 25/50\n",
            "5130/5130 [==============================] - 0s 61us/sample - loss: 0.9235 - acc: 0.6768\n",
            "Epoch 26/50\n",
            "5130/5130 [==============================] - 0s 58us/sample - loss: 0.9349 - acc: 0.6791\n",
            "Epoch 27/50\n",
            "5130/5130 [==============================] - 0s 58us/sample - loss: 0.9060 - acc: 0.6858\n",
            "Epoch 28/50\n",
            "5130/5130 [==============================] - 0s 58us/sample - loss: 0.9087 - acc: 0.6770\n",
            "Epoch 29/50\n",
            "5130/5130 [==============================] - 0s 60us/sample - loss: 0.8854 - acc: 0.6916\n",
            "Epoch 30/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 0.8845 - acc: 0.6893\n",
            "Epoch 31/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 0.8867 - acc: 0.6840\n",
            "Epoch 32/50\n",
            "5130/5130 [==============================] - 0s 59us/sample - loss: 0.8693 - acc: 0.7021\n",
            "Epoch 33/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 0.8715 - acc: 0.7031\n",
            "Epoch 34/50\n",
            "5130/5130 [==============================] - 0s 59us/sample - loss: 0.8624 - acc: 0.7004\n",
            "Epoch 35/50\n",
            "5130/5130 [==============================] - 0s 60us/sample - loss: 0.8657 - acc: 0.6969\n",
            "Epoch 36/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 0.8464 - acc: 0.7025\n",
            "Epoch 37/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 0.8450 - acc: 0.7057\n",
            "Epoch 38/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 0.8477 - acc: 0.7078\n",
            "Epoch 39/50\n",
            "5130/5130 [==============================] - 0s 59us/sample - loss: 0.8340 - acc: 0.7119\n",
            "Epoch 40/50\n",
            "5130/5130 [==============================] - 0s 59us/sample - loss: 0.8413 - acc: 0.7033\n",
            "Epoch 41/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 0.8349 - acc: 0.7092\n",
            "Epoch 42/50\n",
            "5130/5130 [==============================] - 0s 59us/sample - loss: 0.8139 - acc: 0.7164\n",
            "Epoch 43/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 0.8100 - acc: 0.7211\n",
            "Epoch 44/50\n",
            "5130/5130 [==============================] - 0s 56us/sample - loss: 0.8083 - acc: 0.7214\n",
            "Epoch 45/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 0.8056 - acc: 0.7148\n",
            "Epoch 46/50\n",
            "5130/5130 [==============================] - 0s 59us/sample - loss: 0.7976 - acc: 0.7242\n",
            "Epoch 47/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 0.7979 - acc: 0.7156\n",
            "Epoch 48/50\n",
            "5130/5130 [==============================] - 0s 58us/sample - loss: 0.7916 - acc: 0.7220\n",
            "Epoch 49/50\n",
            "5130/5130 [==============================] - 0s 59us/sample - loss: 0.7925 - acc: 0.7183\n",
            "Epoch 50/50\n",
            "5130/5130 [==============================] - 0s 57us/sample - loss: 0.7731 - acc: 0.7349\n",
            "Epoch 1/50\n",
            "5130/5130 [==============================] - 1s 131us/sample - loss: 1.8580 - acc: 0.3405\n",
            "Epoch 2/50\n",
            "5130/5130 [==============================] - 0s 63us/sample - loss: 1.4987 - acc: 0.4778\n",
            "Epoch 3/50\n",
            "5130/5130 [==============================] - 0s 65us/sample - loss: 1.3844 - acc: 0.5160\n",
            "Epoch 4/50\n",
            "5130/5130 [==============================] - 0s 63us/sample - loss: 1.3124 - acc: 0.5279\n",
            "Epoch 5/50\n",
            "5130/5130 [==============================] - 0s 61us/sample - loss: 1.2613 - acc: 0.5583\n",
            "Epoch 6/50\n",
            "5130/5130 [==============================] - 0s 65us/sample - loss: 1.2277 - acc: 0.5630\n",
            "Epoch 7/50\n",
            "5130/5130 [==============================] - 0s 61us/sample - loss: 1.1994 - acc: 0.5749\n",
            "Epoch 8/50\n",
            "5130/5130 [==============================] - 0s 63us/sample - loss: 1.1607 - acc: 0.5899\n",
            "Epoch 9/50\n",
            "5130/5130 [==============================] - 0s 61us/sample - loss: 1.1391 - acc: 0.5926\n",
            "Epoch 10/50\n",
            "5130/5130 [==============================] - 0s 65us/sample - loss: 1.1083 - acc: 0.6058\n",
            "Epoch 11/50\n",
            "5130/5130 [==============================] - 0s 64us/sample - loss: 1.0972 - acc: 0.6105\n",
            "Epoch 12/50\n",
            "5130/5130 [==============================] - 0s 62us/sample - loss: 1.0762 - acc: 0.6201\n",
            "Epoch 13/50\n",
            "5130/5130 [==============================] - 0s 64us/sample - loss: 1.0669 - acc: 0.6226\n",
            "Epoch 14/50\n",
            "5130/5130 [==============================] - 0s 63us/sample - loss: 1.0374 - acc: 0.6287\n",
            "Epoch 15/50\n",
            "5130/5130 [==============================] - 0s 61us/sample - loss: 1.0232 - acc: 0.6407\n",
            "Epoch 16/50\n",
            "5130/5130 [==============================] - 0s 66us/sample - loss: 1.0100 - acc: 0.6468\n",
            "Epoch 17/50\n",
            "5130/5130 [==============================] - 0s 64us/sample - loss: 1.0048 - acc: 0.6378\n",
            "Epoch 18/50\n",
            "5130/5130 [==============================] - 0s 63us/sample - loss: 0.9811 - acc: 0.6575\n",
            "Epoch 19/50\n",
            "5130/5130 [==============================] - 0s 65us/sample - loss: 0.9722 - acc: 0.6593\n",
            "Epoch 20/50\n",
            "5130/5130 [==============================] - 0s 63us/sample - loss: 0.9689 - acc: 0.6616\n",
            "Epoch 21/50\n",
            "5130/5130 [==============================] - 0s 61us/sample - loss: 0.9629 - acc: 0.6608\n",
            "Epoch 22/50\n",
            "5130/5130 [==============================] - 0s 65us/sample - loss: 0.9523 - acc: 0.6667\n",
            "Epoch 23/50\n",
            "5130/5130 [==============================] - 0s 62us/sample - loss: 0.9401 - acc: 0.6708\n",
            "Epoch 24/50\n",
            "5130/5130 [==============================] - 0s 63us/sample - loss: 0.9288 - acc: 0.6717\n",
            "Epoch 25/50\n",
            "5130/5130 [==============================] - 0s 65us/sample - loss: 0.9235 - acc: 0.6725\n",
            "Epoch 26/50\n",
            "5130/5130 [==============================] - 0s 60us/sample - loss: 0.9121 - acc: 0.6811\n",
            "Epoch 27/50\n",
            "5130/5130 [==============================] - 0s 65us/sample - loss: 0.9077 - acc: 0.6862\n",
            "Epoch 28/50\n",
            "5130/5130 [==============================] - 0s 65us/sample - loss: 0.8829 - acc: 0.6873\n",
            "Epoch 29/50\n",
            "5130/5130 [==============================] - 0s 61us/sample - loss: 0.8829 - acc: 0.6918\n",
            "Epoch 30/50\n",
            "5130/5130 [==============================] - 0s 62us/sample - loss: 0.8816 - acc: 0.6867\n",
            "Epoch 31/50\n",
            "5130/5130 [==============================] - 0s 67us/sample - loss: 0.8785 - acc: 0.6906\n",
            "Epoch 32/50\n",
            "5130/5130 [==============================] - 0s 63us/sample - loss: 0.8764 - acc: 0.6955\n",
            "Epoch 33/50\n",
            "5130/5130 [==============================] - 0s 62us/sample - loss: 0.8562 - acc: 0.7047\n",
            "Epoch 34/50\n",
            "5130/5130 [==============================] - 0s 66us/sample - loss: 0.8661 - acc: 0.6977\n",
            "Epoch 35/50\n",
            "5130/5130 [==============================] - 0s 62us/sample - loss: 0.8625 - acc: 0.6979\n",
            "Epoch 36/50\n",
            "5130/5130 [==============================] - 0s 62us/sample - loss: 0.8455 - acc: 0.6981\n",
            "Epoch 37/50\n",
            "5130/5130 [==============================] - 0s 61us/sample - loss: 0.8426 - acc: 0.7072\n",
            "Epoch 38/50\n",
            "5130/5130 [==============================] - 0s 65us/sample - loss: 0.8326 - acc: 0.7064\n",
            "Epoch 39/50\n",
            "5130/5130 [==============================] - 0s 62us/sample - loss: 0.8274 - acc: 0.7090\n",
            "Epoch 40/50\n",
            "5130/5130 [==============================] - 0s 61us/sample - loss: 0.8305 - acc: 0.7175\n",
            "Epoch 41/50\n",
            "5130/5130 [==============================] - 0s 66us/sample - loss: 0.8343 - acc: 0.7107\n",
            "Epoch 42/50\n",
            "5130/5130 [==============================] - 0s 63us/sample - loss: 0.8137 - acc: 0.7207\n",
            "Epoch 43/50\n",
            "5130/5130 [==============================] - 0s 63us/sample - loss: 0.8167 - acc: 0.7148\n",
            "Epoch 44/50\n",
            "5130/5130 [==============================] - 0s 65us/sample - loss: 0.7996 - acc: 0.7216\n",
            "Epoch 45/50\n",
            "5130/5130 [==============================] - 0s 63us/sample - loss: 0.8095 - acc: 0.7187\n",
            "Epoch 46/50\n",
            "5130/5130 [==============================] - 0s 62us/sample - loss: 0.7933 - acc: 0.7238\n",
            "Epoch 47/50\n",
            "5130/5130 [==============================] - 0s 65us/sample - loss: 0.7854 - acc: 0.7298\n",
            "Epoch 48/50\n",
            "5130/5130 [==============================] - 0s 62us/sample - loss: 0.7991 - acc: 0.7240\n",
            "Epoch 49/50\n",
            "5130/5130 [==============================] - 0s 61us/sample - loss: 0.7809 - acc: 0.7310\n",
            "Epoch 50/50\n",
            "5130/5130 [==============================] - 0s 63us/sample - loss: 0.7763 - acc: 0.7261\n",
            "Epoch 1/50\n",
            "5700/5700 [==============================] - 2s 314us/sample - loss: 1.6271 - acc: 0.4330\n",
            "Epoch 2/50\n",
            "5700/5700 [==============================] - 1s 247us/sample - loss: 1.3266 - acc: 0.5235\n",
            "Epoch 3/50\n",
            "5700/5700 [==============================] - 1s 250us/sample - loss: 1.2194 - acc: 0.5568\n",
            "Epoch 4/50\n",
            "5700/5700 [==============================] - 1s 246us/sample - loss: 1.1725 - acc: 0.5782\n",
            "Epoch 5/50\n",
            "5700/5700 [==============================] - 1s 246us/sample - loss: 1.1205 - acc: 0.5958\n",
            "Epoch 6/50\n",
            "5700/5700 [==============================] - 1s 246us/sample - loss: 1.0896 - acc: 0.6088\n",
            "Epoch 7/50\n",
            "5700/5700 [==============================] - 1s 249us/sample - loss: 1.0543 - acc: 0.6249\n",
            "Epoch 8/50\n",
            "5700/5700 [==============================] - 1s 248us/sample - loss: 1.0190 - acc: 0.6382\n",
            "Epoch 9/50\n",
            "5700/5700 [==============================] - 1s 247us/sample - loss: 1.0067 - acc: 0.6468\n",
            "Epoch 10/50\n",
            "5700/5700 [==============================] - 1s 246us/sample - loss: 0.9776 - acc: 0.6525\n",
            "Epoch 11/50\n",
            "5700/5700 [==============================] - 1s 245us/sample - loss: 0.9593 - acc: 0.6595\n",
            "Epoch 12/50\n",
            "5700/5700 [==============================] - 1s 249us/sample - loss: 0.9490 - acc: 0.6593\n",
            "Epoch 13/50\n",
            "5700/5700 [==============================] - 1s 244us/sample - loss: 0.9177 - acc: 0.6761\n",
            "Epoch 14/50\n",
            "5700/5700 [==============================] - 1s 248us/sample - loss: 0.9176 - acc: 0.6691\n",
            "Epoch 15/50\n",
            "5700/5700 [==============================] - 1s 245us/sample - loss: 0.8993 - acc: 0.6793\n",
            "Epoch 16/50\n",
            "5700/5700 [==============================] - 1s 248us/sample - loss: 0.8903 - acc: 0.6840\n",
            "Epoch 17/50\n",
            "5700/5700 [==============================] - 1s 250us/sample - loss: 0.8812 - acc: 0.6909\n",
            "Epoch 18/50\n",
            "5700/5700 [==============================] - 1s 243us/sample - loss: 0.8709 - acc: 0.6877\n",
            "Epoch 19/50\n",
            "5700/5700 [==============================] - 1s 246us/sample - loss: 0.8433 - acc: 0.7032\n",
            "Epoch 20/50\n",
            "5700/5700 [==============================] - 1s 245us/sample - loss: 0.8448 - acc: 0.7040\n",
            "Epoch 21/50\n",
            "5700/5700 [==============================] - 1s 245us/sample - loss: 0.8365 - acc: 0.7086\n",
            "Epoch 22/50\n",
            "5700/5700 [==============================] - 1s 249us/sample - loss: 0.8198 - acc: 0.7126\n",
            "Epoch 23/50\n",
            "5700/5700 [==============================] - 1s 246us/sample - loss: 0.8203 - acc: 0.7135\n",
            "Epoch 24/50\n",
            "5700/5700 [==============================] - 1s 243us/sample - loss: 0.8044 - acc: 0.7172\n",
            "Epoch 25/50\n",
            "5700/5700 [==============================] - 1s 242us/sample - loss: 0.7926 - acc: 0.7161\n",
            "Epoch 26/50\n",
            "5700/5700 [==============================] - 1s 244us/sample - loss: 0.7887 - acc: 0.7212\n",
            "Epoch 27/50\n",
            "5700/5700 [==============================] - 1s 246us/sample - loss: 0.7720 - acc: 0.7309\n",
            "Epoch 28/50\n",
            "5700/5700 [==============================] - 1s 250us/sample - loss: 0.7701 - acc: 0.7328\n",
            "Epoch 29/50\n",
            "5700/5700 [==============================] - 1s 247us/sample - loss: 0.7677 - acc: 0.7246\n",
            "Epoch 30/50\n",
            "5700/5700 [==============================] - 1s 250us/sample - loss: 0.7561 - acc: 0.7330\n",
            "Epoch 31/50\n",
            "5700/5700 [==============================] - 1s 248us/sample - loss: 0.7548 - acc: 0.7375\n",
            "Epoch 32/50\n",
            "5700/5700 [==============================] - 1s 248us/sample - loss: 0.7439 - acc: 0.7425\n",
            "Epoch 33/50\n",
            "5700/5700 [==============================] - 1s 251us/sample - loss: 0.7287 - acc: 0.7440\n",
            "Epoch 34/50\n",
            "5700/5700 [==============================] - 1s 248us/sample - loss: 0.7371 - acc: 0.7396\n",
            "Epoch 35/50\n",
            "5700/5700 [==============================] - 1s 244us/sample - loss: 0.7199 - acc: 0.7446\n",
            "Epoch 36/50\n",
            "5700/5700 [==============================] - 1s 248us/sample - loss: 0.7127 - acc: 0.7502\n",
            "Epoch 37/50\n",
            "5700/5700 [==============================] - 1s 244us/sample - loss: 0.7008 - acc: 0.7563\n",
            "Epoch 38/50\n",
            "5700/5700 [==============================] - 1s 249us/sample - loss: 0.7029 - acc: 0.7461\n",
            "Epoch 39/50\n",
            "5700/5700 [==============================] - 1s 242us/sample - loss: 0.7124 - acc: 0.7426\n",
            "Epoch 40/50\n",
            "5700/5700 [==============================] - 1s 246us/sample - loss: 0.6993 - acc: 0.7551\n",
            "Epoch 41/50\n",
            "5700/5700 [==============================] - 1s 246us/sample - loss: 0.6905 - acc: 0.7525\n",
            "Epoch 42/50\n",
            "5700/5700 [==============================] - 1s 246us/sample - loss: 0.6838 - acc: 0.7649\n",
            "Epoch 43/50\n",
            "5700/5700 [==============================] - 1s 245us/sample - loss: 0.6833 - acc: 0.7626\n",
            "Epoch 44/50\n",
            "5700/5700 [==============================] - 1s 250us/sample - loss: 0.6734 - acc: 0.7619\n",
            "Epoch 45/50\n",
            "5700/5700 [==============================] - 1s 245us/sample - loss: 0.6698 - acc: 0.7618\n",
            "Epoch 46/50\n",
            "5700/5700 [==============================] - 1s 244us/sample - loss: 0.6705 - acc: 0.7588\n",
            "Epoch 47/50\n",
            "5700/5700 [==============================] - 1s 244us/sample - loss: 0.6509 - acc: 0.7737\n",
            "Epoch 48/50\n",
            "5700/5700 [==============================] - 1s 249us/sample - loss: 0.6575 - acc: 0.7709\n",
            "Epoch 49/50\n",
            "5700/5700 [==============================] - 1s 247us/sample - loss: 0.6566 - acc: 0.7725\n",
            "Epoch 50/50\n",
            "5700/5700 [==============================] - 1s 248us/sample - loss: 0.6566 - acc: 0.7698\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XfM2equdM-Ro",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "8a9fad94-baf0-454b-c4b6-d86ce3eea89a"
      },
      "source": [
        "print(\"Best: %f using %s\" % (grid_result.best_score_, grid_result.best_params_))\n"
      ],
      "execution_count": 17,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Best: 0.755136 using {'batch_size': 16, 'epochs': 50, 'init': 'normal', 'optimizer': 'adam'}\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-zleafwhyrIn",
        "colab_type": "text"
      },
      "source": [
        "# Демонстрация работы классификатора"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "TCGpkyhmUCZq",
        "colab_type": "text"
      },
      "source": [
        "## Применение нейросети для звукового файла произвольной длины"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Si39jFh9y9vL",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def detect_file_genre(demo_filename):\n",
        "    print(f\"--- {demo_filename} ---\")\n",
        "    # получаем звуковые параметры на 5-секундных интервалах\n",
        "    demo_features = ['chroma_stft', \n",
        "                'rmse',\n",
        "                'tonnetz',\n",
        "                'spectral_centroid', \n",
        "                'spectral_bandwidth', \n",
        "                'spectral_flatness', \n",
        "                'rolloff', \n",
        "                'zero_crossing_rate', \n",
        "                'chroma_cqt',\n",
        "                'melspectrogram',\n",
        "                'tempogram',\n",
        "                'mfcc']\n",
        "\n",
        "    duration = librosa.get_duration(filename = demo_filename)\n",
        "    intervals = range(0, int(duration // slice_duration))\n",
        "    intervals_count = len(intervals)\n",
        "    print(f\"Duration {duration} sec: {intervals_count} intervals\")\n",
        "    X = []\n",
        "    for i in intervals:\n",
        "        print(\"\\r\", end = f'Interval {i + 1} of {intervals_count}')\n",
        "        y, sr = librosa.load(demo_filename, mono=True, offset = i * slice_duration, duration=slice_duration)\n",
        "        row = get_features(y, sr, demo_features)\n",
        "        X.append(row)\n",
        "\n",
        "    # нормализуем теми же значениями, что и тренировочный набор данных\n",
        "    X = scaler.transform(X)\n",
        "\n",
        "    # классифицируем\n",
        "    prediction = model.predict(X)\n",
        "    prediction = np.argmax(prediction, axis=1)\n",
        "\n",
        "    # разбираем результат\n",
        "    unique, counts = np.unique(prediction, return_counts = True)\n",
        "\n",
        "    result = list(zip(-counts, -unique))\n",
        "    result = -np.sort(result, axis=0)\n",
        "    print(f\"\\n\\n--- Classification ---\")\n",
        "    for r in result:\n",
        "        sure = r[0] * 100 // intervals_count\n",
        "        genre = genres[r[1]]\n",
        "        print(f\"{sure}% – {genre}\")\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "VbdVroC-ULjs",
        "colab_type": "text"
      },
      "source": [
        "## Примеры классификации"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "9Z2m8UY5LXGf",
        "colab_type": "text"
      },
      "source": [
        "### Sabaton – Primo Victoria"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xa2NDqwq558B",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 204
        },
        "outputId": "9019da07-af3f-4b41-e4d1-15f12fa444a3"
      },
      "source": [
        "detect_file_genre(\"Sabaton  - Primo Victoria.mp3\")"
      ],
      "execution_count": 113,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--- Sabaton  - Primo Victoria.mp3 ---\n",
            "Duration 272.3 sec: 54 intervals\n",
            "Interval 54 of 54\n",
            "\n",
            "--- Classification ---\n",
            "42% – rock\n",
            "24% – metal\n",
            "22% – hiphop\n",
            "5% – disco\n",
            "3% – country\n",
            "1% – classical\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "O6QJaXpCN1c0",
        "colab_type": "text"
      },
      "source": [
        "### Louis Armstrong – Go Down Moses"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nSIputOSN3lH",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 221
        },
        "outputId": "1e1196cc-0fa2-4e39-94e3-232f0a998adc"
      },
      "source": [
        "detect_file_genre(\"Louis Armstrong-Go Down Moses.mp3\")"
      ],
      "execution_count": 109,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--- Louis Armstrong-Go Down Moses.mp3 ---\n",
            "Duration 220.8 sec: 44 intervals\n",
            "Interval 44 of 44\n",
            "\n",
            "--- Classification ---\n",
            "29% – rock\n",
            "27% – reggae\n",
            "15% – jazz\n",
            "9% – hiphop\n",
            "6% – disco\n",
            "6% – country\n",
            "4% – classical\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "atW3M8HQPCL9",
        "colab_type": "text"
      },
      "source": [
        "### AC/DC – Thunderstuck"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "StS9JVagPFJj",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 187
        },
        "outputId": "bbbf7392-c2c4-4aa2-f5b8-f53d6a953083"
      },
      "source": [
        "detect_file_genre(\"ACDC - Thunderstruck.mp3\")"
      ],
      "execution_count": 114,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--- ACDC - Thunderstruck.mp3 ---\n",
            "Duration 174.7 sec: 34 intervals\n",
            "Interval 34 of 34\n",
            "\n",
            "--- Classification ---\n",
            "35% – rock\n",
            "23% – reggae\n",
            "20% – metal\n",
            "17% – hiphop\n",
            "2% – country\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "s7xMeUOXN8iW",
        "colab_type": "text"
      },
      "source": [
        "### Sting - Englishman In New York"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4npEVmo9OACx",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 238
        },
        "outputId": "5c41a5a8-807a-4255-af61-b7114d9d534a"
      },
      "source": [
        "detect_file_genre(\"Sting - Englishman In New York.mp3\")"
      ],
      "execution_count": 110,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--- Sting - Englishman In New York.mp3 ---\n",
            "Duration 267.8 sec: 53 intervals\n",
            "Interval 53 of 53\n",
            "\n",
            "--- Classification ---\n",
            "32% – rock\n",
            "24% – reggae\n",
            "13% – jazz\n",
            "9% – hiphop\n",
            "9% – disco\n",
            "5% – country\n",
            "3% – classical\n",
            "1% – blues\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "yoX18YEwVkK8",
        "colab_type": "text"
      },
      "source": [
        "### Adele – Skyfall"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "inlPoeKTVmAy",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 204
        },
        "outputId": "64e91ec8-ba75-461c-d00b-7461ca5f59cd"
      },
      "source": [
        "detect_file_genre(\"Adele - Skyfall.mp3\")"
      ],
      "execution_count": 127,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--- Adele - Skyfall.mp3 ---\n",
            "Duration 289.4 sec: 57 intervals\n",
            "Interval 57 of 57\n",
            "\n",
            "--- Classification ---\n",
            "36% – rock\n",
            "33% – reggae\n",
            "14% – pop\n",
            "12% – hiphop\n",
            "1% – country\n",
            "1% – classical\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "dg8YnhE5P-qd",
        "colab_type": "text"
      },
      "source": [
        "### Harry Belafonte – Banana Boat Song"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Qgc55cAAQCvu",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 204
        },
        "outputId": "b2263ab6-7cc2-4544-de26-f2f887a6c541"
      },
      "source": [
        "detect_file_genre(\"Harry Belafonte - Banana Boat Song.mp3\")"
      ],
      "execution_count": 115,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--- Harry Belafonte - Banana Boat Song.mp3 ---\n",
            "Duration 187.3 sec: 37 intervals\n",
            "Interval 37 of 37\n",
            "\n",
            "--- Classification ---\n",
            "35% – reggae\n",
            "21% – pop\n",
            "18% – jazz\n",
            "10% – hiphop\n",
            "10% – country\n",
            "2% – classical\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "MoGPkSRsS0iI",
        "colab_type": "text"
      },
      "source": [
        "### Metallica – Low Man's Lyric"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hgfyiZ7JS5jf",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 204
        },
        "outputId": "5551c0de-bfa6-4b47-cd12-5a06d506e227"
      },
      "source": [
        "detect_file_genre(\"Metallica - Low Mans Lyric.mp3\")"
      ],
      "execution_count": 119,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--- Metallica - Low Mans Lyric.mp3 ---\n",
            "Duration 456.9 sec: 91 intervals\n",
            "Interval 91 of 91\n",
            "\n",
            "--- Classification ---\n",
            "47% – rock\n",
            "24% – metal\n",
            "9% – hiphop\n",
            "9% – country\n",
            "7% – classical\n",
            "1% – blues\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "GVdV0d5sQnXM",
        "colab_type": "text"
      },
      "source": [
        "### Rapsody – Prince Igor"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3o8cOU4PQppT",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 187
        },
        "outputId": "8eb522cc-6d78-4ce1-c741-d95f166841f9"
      },
      "source": [
        "detect_file_genre(\"Rapsody - Prince Igor.mp3\")"
      ],
      "execution_count": 116,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--- Rapsody - Prince Igor.mp3 ---\n",
            "Duration 231.9 sec: 46 intervals\n",
            "Interval 46 of 46\n",
            "\n",
            "--- Classification ---\n",
            "52% – reggae\n",
            "19% – jazz\n",
            "17% – hiphop\n",
            "6% – disco\n",
            "4% – classical\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "m-M40h-XRJg4",
        "colab_type": "text"
      },
      "source": [
        "### Weird Al Yankovic – White and Nerdy"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "B8YnuDnaRQTk",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 153
        },
        "outputId": "a8ce5fa6-bef0-4eca-db8a-107aa365fef1"
      },
      "source": [
        "detect_file_genre(\"Weird Al Yankovic - White  Nerdy.mp3\")"
      ],
      "execution_count": 117,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--- Weird Al Yankovic - White  Nerdy.mp3 ---\n",
            "Duration 170.5 sec: 34 intervals\n",
            "Interval 34 of 34\n",
            "\n",
            "--- Classification ---\n",
            "61% – hiphop\n",
            "23% – disco\n",
            "14% – country\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "CsSLXQ8pLMOl",
        "colab_type": "text"
      },
      "source": [
        "### Infected Mushroom - Artillery"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zajtA3USLObb",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 204
        },
        "outputId": "d22c7d1c-80d8-4789-9707-8254df1698e6"
      },
      "source": [
        "detect_file_genre(\"Infected Mushroom - Artillery.mp3\")"
      ],
      "execution_count": 112,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--- Infected Mushroom - Artillery.mp3 ---\n",
            "Duration 269.8 sec: 53 intervals\n",
            "Interval 53 of 53\n",
            "\n",
            "--- Classification ---\n",
            "32% – rock\n",
            "32% – reggae\n",
            "18% – pop\n",
            "9% – hiphop\n",
            "5% – country\n",
            "1% – blues\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "x23wmJoBLSRc",
        "colab_type": "text"
      },
      "source": [
        "### Katzenjammer - Rock-Paper-Scissors"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XGUfs4edLaes",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 221
        },
        "outputId": "3b0915ea-6be3-42e1-b928-3bcbd816a08f"
      },
      "source": [
        "detect_file_genre(\"Katzenjammer - Rock-Paper-Scissors.mp3\")"
      ],
      "execution_count": 111,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--- Katzenjammer - Rock-Paper-Scissors.mp3 ---\n",
            "Duration 208.7 sec: 41 intervals\n",
            "Interval 41 of 41\n",
            "\n",
            "--- Classification ---\n",
            "39% – reggae\n",
            "29% – pop\n",
            "19% – jazz\n",
            "4% – hiphop\n",
            "2% – disco\n",
            "2% – country\n",
            "2% – classical\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Ly61HzBKXxHh",
        "colab_type": "text"
      },
      "source": [
        "### Boney M – Sunny"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "YoPahiuiXzWy",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 187
        },
        "outputId": "ac36a8bb-9882-41c7-c24f-98aaf620f644"
      },
      "source": [
        "detect_file_genre(\"Boney M - Sunny.mp3\")"
      ],
      "execution_count": 131,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--- Boney M - Sunny.mp3 ---\n",
            "Duration 264.7 sec: 52 intervals\n",
            "Interval 52 of 52\n",
            "\n",
            "--- Classification ---\n",
            "48% – rock\n",
            "23% – jazz\n",
            "23% – hiphop\n",
            "3% – disco\n",
            "1% – classical\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "mEoyeCuUZ_Ii",
        "colab_type": "text"
      },
      "source": [
        "### Lady Gaga – Paparazzi"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "J1D_mWiyaBr0",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 170
        },
        "outputId": "5647213b-2820-468e-b78d-b8007aaf28f3"
      },
      "source": [
        "detect_file_genre(\"Lady Gaga - Paparazzi.mp3\")"
      ],
      "execution_count": 135,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--- Lady Gaga - Paparazzi.mp3 ---\n",
            "Duration 205.1 sec: 41 intervals\n",
            "Interval 41 of 41\n",
            "\n",
            "--- Classification ---\n",
            "68% – rock\n",
            "19% – pop\n",
            "9% – metal\n",
            "2% – hiphop\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "B0aWBFFUagqP",
        "colab_type": "text"
      },
      "source": [
        "### Kiesza – Hideaway"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EdfTJ3wAaldz",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 221
        },
        "outputId": "eceb48d9-9271-4849-8150-d1eebb271cc4"
      },
      "source": [
        "detect_file_genre(\"Kiesza - Hideaway.mp3\")"
      ],
      "execution_count": 136,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--- Kiesza - Hideaway.mp3 ---\n",
            "Duration 274.8 sec: 54 intervals\n",
            "Interval 54 of 54\n",
            "\n",
            "--- Classification ---\n",
            "46% – reggae\n",
            "22% – pop\n",
            "12% – jazz\n",
            "11% – hiphop\n",
            "3% – disco\n",
            "1% – country\n",
            "1% – classical\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "K7SdSXO7cpCm",
        "colab_type": "text"
      },
      "source": [
        "### Ylvis – What Does the Fox Say"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "uCyugEHqcvI-",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 187
        },
        "outputId": "5c9845c0-ed41-47e2-f8e2-7edab49c6cf8"
      },
      "source": [
        "detect_file_genre(\"Ylvis - The Fox.mp3\")"
      ],
      "execution_count": 143,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--- Ylvis - The Fox.mp3 ---\n",
            "Duration 224.8 sec: 44 intervals\n",
            "Interval 44 of 44\n",
            "\n",
            "--- Classification ---\n",
            "65% – pop\n",
            "11% – hiphop\n",
            "9% – disco\n",
            "6% – country\n",
            "6% – classical\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "K9SYqFpUd_be",
        "colab_type": "text"
      },
      "source": [
        "### Lindsey Stirling – Shadows"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "thCksMlOeDPa",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 221
        },
        "outputId": "3b03dc59-9065-4382-9f4c-a76da64f53d3"
      },
      "source": [
        "detect_file_genre(\"Shadows - Lindsey Stirling.mp3\")"
      ],
      "execution_count": 145,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--- Shadows - Lindsey Stirling.mp3 ---\n",
            "Duration 234.6 sec: 46 intervals\n",
            "Interval 46 of 46\n",
            "\n",
            "--- Classification ---\n",
            "56% – rock\n",
            "15% – reggae\n",
            "15% – pop\n",
            "4% – hiphop\n",
            "4% – disco\n",
            "2% – country\n",
            "2% – classical\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "VVU2SiIPR-XC",
        "colab_type": "text"
      },
      "source": [
        "### Зоопарк – Я забываю"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "oomne6XOSBSZ",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 204
        },
        "outputId": "c23d5206-d6bb-4f39-fb7f-c8265462dcda"
      },
      "source": [
        "detect_file_genre(\"Я забываю Зоопарк.mp3\")"
      ],
      "execution_count": 118,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--- Я забываю Зоопарк.mp3 ---\n",
            "Duration 209.9 sec: 41 intervals\n",
            "Interval 41 of 41\n",
            "\n",
            "--- Classification ---\n",
            "36% – rock\n",
            "29% – metal\n",
            "24% – jazz\n",
            "4% – hiphop\n",
            "2% – disco\n",
            "2% – country\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "CR6xcttQTswP",
        "colab_type": "text"
      },
      "source": [
        "### Нейромонах Феофан – Драм для Ивана"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DHEybp5ITw08",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 238
        },
        "outputId": "26dbec53-4339-47f9-94f1-474f73cde370"
      },
      "source": [
        "detect_file_genre(\"Драм для Ивана - Нейромонах Феофан.mp3\")"
      ],
      "execution_count": 123,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--- Драм для Ивана - Нейромонах Феофан.mp3 ---\n",
            "Duration 252.9 sec: 50 intervals\n",
            "Interval 50 of 50\n",
            "\n",
            "--- Classification ---\n",
            "58% – rock\n",
            "8% – reggae\n",
            "8% – pop\n",
            "6% – jazz\n",
            "6% – hiphop\n",
            "6% – disco\n",
            "4% – country\n",
            "4% – blues\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "TifwRYxoWqRo",
        "colab_type": "text"
      },
      "source": [
        "### 5'nizza – Пятница"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hx842of2WuR0",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 187
        },
        "outputId": "748a7e8c-e8ec-4874-e3ae-da79392f570d"
      },
      "source": [
        "detect_file_genre(\"5nizza - Пятница.mp3\")"
      ],
      "execution_count": 128,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--- 5nizza - Пятница.mp3 ---\n",
            "Duration 178.9 sec: 35 intervals\n",
            "Interval 35 of 35\n",
            "\n",
            "--- Classification ---\n",
            "60% – rock\n",
            "17% – reggae\n",
            "17% – jazz\n",
            "2% – hiphop\n",
            "2% – country\n"
          ],
          "name": "stdout"
        }
      ]
    }
  ]
}