cnn-in-digit-recognition-99-5.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "881a04a4",
   "metadata": {},
   "source": [
    "<a id=\"1\"></a>\n",
    "<h1 style=\"text-align: center; \n",
    "           border-radius: 15px; \n",
    "           padding: 20px; \n",
    "           margin: 0; \n",
    "           font-size: 180%; \n",
    "           font-family: Arial, sans-serif; \n",
    "           background-color: #7A1330; \n",
    "           color: #F5F5F5; \n",
    "           box-shadow: 0 6px 12px rgba(0, 0, 0, 0.2); \n",
    "           border: 2px solid #5E0E23;\">\n",
    "   <b>Power of Neural Networks in Digit Recognition</b>\n",
    "</h1>"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "221d40ca",
   "metadata": {
    "papermill": {
     "duration": 0.010134,
     "end_time": "2024-09-23T20:30:53.925889",
     "exception": false,
     "start_time": "2024-09-23T20:30:53.915755",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "![ai](https://r2.starryai.com/results/1050111228/f615547a-0894-401a-8e47-ec9a50dd8ed7.webp)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "473df01e",
   "metadata": {},
   "source": [
    "<h2 style=\"text-align: center; \n",
    "           border-radius: 12px; \n",
    "           padding: 15px; \n",
    "           margin: 0; \n",
    "           font-size: 140%; \n",
    "           font-family: Arial, sans-serif; \n",
    "           background-color: #046B85; \n",
    "           color: #F5F5F5; \n",
    "           box-shadow: 0 3px 6px rgba(0, 0, 0, 0.2); \n",
    "           border: 1px solid #034C63;\">\n",
    "   <b>1. Introduction</b>\n",
    "</h2>"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "58b96c3e",
   "metadata": {
    "papermill": {
     "duration": 0.01007,
     "end_time": "2024-09-23T20:30:53.966818",
     "exception": false,
     "start_time": "2024-09-23T20:30:53.956748",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "In the rapidly evolving field of artificial intelligence (AI), neural networks have become a powerful tool for solving complex problems, such as digit recognition. This project begins by establishing a baseline using dense neural networks, before transitioning to more advanced convolutional neural networks (CNNs). The aim is to showcase how different types of neural networks can be applied to the MNIST dataset, which contains 28,000 images of handwritten digits, and progressively improve accuracy in digit classification.\n",
    "\n",
    "### Understanding Neural Networks\n",
    "Neural networks are inspired by the structure and function of the human brain. Just as neurons in the brain process information, artificial neurons in a neural network take inputs, process them, and pass on the result. These neurons are organized into layers, where each layer learns progressively more complex patterns in the data.\n",
    "\n",
    "We begin this project with dense (fully connected) neural networks, where each neuron in one layer is connected to every neuron in the next. These layers extract features from the input data (in our case, the pixel values of handwritten digits) and perform non-linear transformations to predict the digit. Dense networks are effective for simple datasets, but they often struggle with image data, where spatial relationships between pixels are key.\n",
    "\n",
    "### From Dense Networks to CNNs\n",
    "While dense neural networks can serve as a strong baseline, convolutional neural networks (CNNs) are far better suited for image classification tasks. CNNs take advantage of the spatial structure of images by applying filters (convolutions) to detect patterns such as edges, curves, and textures, allowing the network to \"see\" the image in a more structured way. These filters automatically learn to detect features at different levels of abstraction, making CNNs more effective in recognizing and classifying complex patterns in images, such as handwritten digits."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "42744cd8",
   "metadata": {
    "papermill": {
     "duration": 0.010266,
     "end_time": "2024-09-23T20:30:53.987399",
     "exception": false,
     "start_time": "2024-09-23T20:30:53.977133",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "<h2 style=\"text-align: center; \n",
    "           border-radius: 12px; \n",
    "           padding: 15px; \n",
    "           margin: 0; \n",
    "           font-size: 120%; \n",
    "           font-family: Arial, sans-serif; \n",
    "           background-color: #046B85; \n",
    "           color: #F5F5F5; \n",
    "           box-shadow: 0 3px 6px rgba(0, 0, 0, 0.2); \n",
    "           border: 1px solid #034C63;\">\n",
    "   <b>2. Importing Libraries</b>\n",
    "</h2>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "0ef38efc",
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2024-09-23T20:30:54.010311Z",
     "iopub.status.busy": "2024-09-23T20:30:54.009505Z",
     "iopub.status.idle": "2024-09-23T20:31:09.906246Z",
     "shell.execute_reply": "2024-09-23T20:31:09.905464Z"
    },
    "papermill": {
     "duration": 15.910789,
     "end_time": "2024-09-23T20:31:09.908790",
     "exception": false,
     "start_time": "2024-09-23T20:30:53.998001",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "import numpy as np \n",
    "import pandas as pd \n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "from sklearn.metrics import confusion_matrix\n",
    "\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "import tensorflow as tf\n",
    "from tensorflow.keras import Sequential\n",
    "from tensorflow.keras.layers import Conv2D, MaxPooling2D, Flatten, Dense, Dropout, BatchNormalization\n",
    "from tensorflow.keras.regularizers import L2\n",
    "from tensorflow.keras.losses import SparseCategoricalCrossentropy\n",
    "from tensorflow.keras.metrics import SparseCategoricalAccuracy\n",
    "from tensorflow.keras.optimizers import Adam\n",
    "from tensorflow.keras.preprocessing.image import ImageDataGenerator\n",
    "\n",
    "# Suppress warnings\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2a0e2091",
   "metadata": {
    "papermill": {
     "duration": 0.0104,
     "end_time": "2024-09-23T20:31:09.930190",
     "exception": false,
     "start_time": "2024-09-23T20:31:09.919790",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "<h2 style=\"text-align: center; \n",
    "           border-radius: 12px; \n",
    "           padding: 15px; \n",
    "           margin: 0; \n",
    "           font-size: 120%; \n",
    "           font-family: Arial, sans-serif; \n",
    "           background-color: #046B85; \n",
    "           color: #F5F5F5; \n",
    "           box-shadow: 0 3px 6px rgba(0, 0, 0, 0.2); \n",
    "           border: 1px solid #034C63;\">\n",
    "   <b>3. Data Preparation</b>\n",
    "</h2>"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7e3ee3a2",
   "metadata": {
    "papermill": {
     "duration": 0.01067,
     "end_time": "2024-09-23T20:31:09.951286",
     "exception": false,
     "start_time": "2024-09-23T20:31:09.940616",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "### Loading data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e27947f0",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-09-23T20:31:09.975536Z",
     "iopub.status.busy": "2024-09-23T20:31:09.974894Z",
     "iopub.status.idle": "2024-09-23T20:31:15.368447Z",
     "shell.execute_reply": "2024-09-23T20:31:15.367660Z"
    },
    "papermill": {
     "duration": 5.408337,
     "end_time": "2024-09-23T20:31:15.370824",
     "exception": false,
     "start_time": "2024-09-23T20:31:09.962487",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "train = pd.read_csv(\"train.csv\")\n",
    "test = pd.read_csv('test.csv')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6a2dff3f",
   "metadata": {
    "papermill": {
     "duration": 0.010421,
     "end_time": "2024-09-23T20:31:15.392234",
     "exception": false,
     "start_time": "2024-09-23T20:31:15.381813",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "### Separating features and labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "3c393ae8",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-09-23T20:31:15.414916Z",
     "iopub.status.busy": "2024-09-23T20:31:15.414130Z",
     "iopub.status.idle": "2024-09-23T20:31:15.506921Z",
     "shell.execute_reply": "2024-09-23T20:31:15.506136Z"
    },
    "papermill": {
     "duration": 0.106726,
     "end_time": "2024-09-23T20:31:15.509370",
     "exception": false,
     "start_time": "2024-09-23T20:31:15.402644",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "X = train.drop(\"label\", axis=1) \n",
    "y = train.label"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7330c15a",
   "metadata": {
    "papermill": {
     "duration": 0.010103,
     "end_time": "2024-09-23T20:31:15.530352",
     "exception": false,
     "start_time": "2024-09-23T20:31:15.520249",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "### Data normalization"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c898b9f1",
   "metadata": {
    "papermill": {
     "duration": 0.009979,
     "end_time": "2024-09-23T20:31:15.550617",
     "exception": false,
     "start_time": "2024-09-23T20:31:15.540638",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "This step ensures that our input data is in a suitable range for training a neural network, leading to better performance and faster learning."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "89c3bf99",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-09-23T20:31:15.572615Z",
     "iopub.status.busy": "2024-09-23T20:31:15.572006Z",
     "iopub.status.idle": "2024-09-23T20:31:15.664694Z",
     "shell.execute_reply": "2024-09-23T20:31:15.663855Z"
    },
    "papermill": {
     "duration": 0.106147,
     "end_time": "2024-09-23T20:31:15.666986",
     "exception": false,
     "start_time": "2024-09-23T20:31:15.560839",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Normalize the data (scale pixel values between 0 and 1)\n",
    "X = X / 255.0\n",
    "X_test = test / 255.0"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6969bfd2",
   "metadata": {
    "papermill": {
     "duration": 0.010059,
     "end_time": "2024-09-23T20:31:15.687640",
     "exception": false,
     "start_time": "2024-09-23T20:31:15.677581",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "### Spliting data into training and testing sets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "6388b7ae",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-09-23T20:31:15.709686Z",
     "iopub.status.busy": "2024-09-23T20:31:15.708997Z",
     "iopub.status.idle": "2024-09-23T20:31:15.892199Z",
     "shell.execute_reply": "2024-09-23T20:31:15.891313Z"
    },
    "papermill": {
     "duration": 0.196845,
     "end_time": "2024-09-23T20:31:15.894742",
     "exception": false,
     "start_time": "2024-09-23T20:31:15.697897",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "X_train, X_val, y_train, y_val = train_test_split(X, y, test_size=0.1, random_state=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "300a5253",
   "metadata": {
    "papermill": {
     "duration": 0.010056,
     "end_time": "2024-09-23T20:31:15.916626",
     "exception": false,
     "start_time": "2024-09-23T20:31:15.906570",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "### Preview Images"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "13bae7a3",
   "metadata": {
    "_kg_hide-input": true,
    "execution": {
     "iopub.execute_input": "2024-09-23T20:31:15.939021Z",
     "iopub.status.busy": "2024-09-23T20:31:15.938368Z",
     "iopub.status.idle": "2024-09-23T20:31:17.066339Z",
     "shell.execute_reply": "2024-09-23T20:31:17.065506Z"
    },
    "papermill": {
     "duration": 1.141717,
     "end_time": "2024-09-23T20:31:17.068714",
     "exception": false,
     "start_time": "2024-09-23T20:31:15.926997",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1500x500 with 30 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "X_reshaped = X_train.to_numpy().reshape(-1, 28, 28, 1)\n",
    "\n",
    "plt.figure(figsize=(15, 5))\n",
    "for i in range(30):  \n",
    "    plt.subplot(3, 10, i+1)\n",
    "    plt.imshow(X_reshaped[i, :, :, 0], cmap=plt.cm.binary)  # Access the 28x28 image directly\n",
    "    plt.axis('off')\n",
    "plt.subplots_adjust(wspace=-0.1, hspace=-0.1)\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3f0ec368",
   "metadata": {
    "papermill": {
     "duration": 0.010905,
     "end_time": "2024-09-23T20:31:17.090918",
     "exception": false,
     "start_time": "2024-09-23T20:31:17.080013",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "<h2 style=\"text-align: center; \n",
    "           border-radius: 12px; \n",
    "           padding: 15px; \n",
    "           margin: 0; \n",
    "           font-size: 120%; \n",
    "           font-family: Arial, sans-serif; \n",
    "           background-color: #046B85; \n",
    "           color: #F5F5F5; \n",
    "           box-shadow: 0 3px 6px rgba(0, 0, 0, 0.2); \n",
    "           border: 1px solid #034C63;\">\n",
    "   <b>4. Dense Neural Network</b>\n",
    "</h2>"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "62912c5d",
   "metadata": {
    "papermill": {
     "duration": 0.010774,
     "end_time": "2024-09-23T20:31:17.112742",
     "exception": false,
     "start_time": "2024-09-23T20:31:17.101968",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "A dense neural network, also known as a fully connected neural network (FCNN) or multi-layer perceptron (MLP), is a machine learning model that connects each layer to the previous layer."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8e827e10",
   "metadata": {
    "papermill": {
     "duration": 0.010552,
     "end_time": "2024-09-23T20:31:17.134158",
     "exception": false,
     "start_time": "2024-09-23T20:31:17.123606",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "We will train the model through iterative steps, starting with a simple version. If the model shows signs of high bias or underfitting, we'll increase its complexity. If the model shows high variance or overfitting, we'll boost the regularization. This approach should help us achieve better results."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "6e899847",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-09-23T20:31:17.157025Z",
     "iopub.status.busy": "2024-09-23T20:31:17.156628Z",
     "iopub.status.idle": "2024-09-23T20:31:17.167433Z",
     "shell.execute_reply": "2024-09-23T20:31:17.166732Z"
    },
    "papermill": {
     "duration": 0.024354,
     "end_time": "2024-09-23T20:31:17.169311",
     "exception": false,
     "start_time": "2024-09-23T20:31:17.144957",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Define a simple model\n",
    "model = Sequential([\n",
    "    Dense(units=25, activation='relu'),\n",
    "    Dense(units=15, activation='relu'),\n",
    "    Dense(units=10, activation='softmax')\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "b287c2a6",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-09-23T20:31:17.191988Z",
     "iopub.status.busy": "2024-09-23T20:31:17.191677Z",
     "iopub.status.idle": "2024-09-23T20:31:18.040870Z",
     "shell.execute_reply": "2024-09-23T20:31:18.039970Z"
    },
    "papermill": {
     "duration": 0.863079,
     "end_time": "2024-09-23T20:31:18.043245",
     "exception": false,
     "start_time": "2024-09-23T20:31:17.180166",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Compile the model with an Adam optimizer and metrics\n",
    "model.compile(\n",
    "    optimizer = Adam(learning_rate=1e-3),\n",
    "    loss=SparseCategoricalCrossentropy(),\n",
    "    metrics=['accuracy']\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "8d5b3f52",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-09-23T20:31:18.066529Z",
     "iopub.status.busy": "2024-09-23T20:31:18.066197Z",
     "iopub.status.idle": "2024-09-23T20:32:18.657264Z",
     "shell.execute_reply": "2024-09-23T20:32:18.656306Z"
    },
    "papermill": {
     "duration": 60.60492,
     "end_time": "2024-09-23T20:32:18.659328",
     "exception": false,
     "start_time": "2024-09-23T20:31:18.054408",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/30\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING: All log messages before absl::InitializeLog() is called are written to STDERR\n",
      "I0000 00:00:1727123479.464082      71 service.cc:145] XLA service 0x7aefb0016160 initialized for platform CUDA (this does not guarantee that XLA will be used). Devices:\n",
      "I0000 00:00:1727123479.464136      71 service.cc:153]   StreamExecutor device (0): Tesla T4, Compute Capability 7.5\n",
      "I0000 00:00:1727123479.464142      71 service.cc:153]   StreamExecutor device (1): Tesla T4, Compute Capability 7.5\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[1m 114/1182\u001b[0m \u001b[32m━\u001b[0m\u001b[37m━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[1m1s\u001b[0m 1ms/step - accuracy: 0.3487 - loss: 1.9690"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "I0000 00:00:1727123481.388812      71 device_compiler.h:188] Compiled cluster using XLA!  This line is logged at most once for the lifetime of the process.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m6s\u001b[0m 3ms/step - accuracy: 0.7317 - loss: 0.8819 - val_accuracy: 0.9233 - val_loss: 0.2592\n",
      "Epoch 2/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9284 - loss: 0.2408 - val_accuracy: 0.9429 - val_loss: 0.2040\n",
      "Epoch 3/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9435 - loss: 0.1847 - val_accuracy: 0.9450 - val_loss: 0.1825\n",
      "Epoch 4/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9531 - loss: 0.1578 - val_accuracy: 0.9460 - val_loss: 0.1735\n",
      "Epoch 5/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 1ms/step - accuracy: 0.9590 - loss: 0.1375 - val_accuracy: 0.9517 - val_loss: 0.1553\n",
      "Epoch 6/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9629 - loss: 0.1220 - val_accuracy: 0.9540 - val_loss: 0.1572\n",
      "Epoch 7/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9667 - loss: 0.1087 - val_accuracy: 0.9533 - val_loss: 0.1484\n",
      "Epoch 8/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9705 - loss: 0.0996 - val_accuracy: 0.9576 - val_loss: 0.1384\n",
      "Epoch 9/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9742 - loss: 0.0850 - val_accuracy: 0.9571 - val_loss: 0.1428\n",
      "Epoch 10/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 1ms/step - accuracy: 0.9746 - loss: 0.0819 - val_accuracy: 0.9612 - val_loss: 0.1356\n",
      "Epoch 11/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9762 - loss: 0.0755 - val_accuracy: 0.9614 - val_loss: 0.1392\n",
      "Epoch 12/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 1ms/step - accuracy: 0.9781 - loss: 0.0688 - val_accuracy: 0.9624 - val_loss: 0.1388\n",
      "Epoch 13/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 1ms/step - accuracy: 0.9811 - loss: 0.0641 - val_accuracy: 0.9602 - val_loss: 0.1406\n",
      "Epoch 14/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9812 - loss: 0.0602 - val_accuracy: 0.9612 - val_loss: 0.1424\n",
      "Epoch 15/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 1ms/step - accuracy: 0.9820 - loss: 0.0568 - val_accuracy: 0.9602 - val_loss: 0.1447\n",
      "Epoch 16/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 1ms/step - accuracy: 0.9832 - loss: 0.0520 - val_accuracy: 0.9610 - val_loss: 0.1523\n",
      "Epoch 17/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 1ms/step - accuracy: 0.9848 - loss: 0.0486 - val_accuracy: 0.9624 - val_loss: 0.1473\n",
      "Epoch 18/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9848 - loss: 0.0480 - val_accuracy: 0.9645 - val_loss: 0.1453\n",
      "Epoch 19/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9861 - loss: 0.0460 - val_accuracy: 0.9626 - val_loss: 0.1483\n",
      "Epoch 20/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9868 - loss: 0.0423 - val_accuracy: 0.9619 - val_loss: 0.1495\n",
      "Epoch 21/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9873 - loss: 0.0413 - val_accuracy: 0.9590 - val_loss: 0.1637\n",
      "Epoch 22/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9876 - loss: 0.0376 - val_accuracy: 0.9614 - val_loss: 0.1621\n",
      "Epoch 23/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9892 - loss: 0.0336 - val_accuracy: 0.9595 - val_loss: 0.1689\n",
      "Epoch 24/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9890 - loss: 0.0344 - val_accuracy: 0.9626 - val_loss: 0.1675\n",
      "Epoch 25/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9908 - loss: 0.0297 - val_accuracy: 0.9583 - val_loss: 0.1901\n",
      "Epoch 26/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9899 - loss: 0.0314 - val_accuracy: 0.9624 - val_loss: 0.1853\n",
      "Epoch 27/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9915 - loss: 0.0277 - val_accuracy: 0.9552 - val_loss: 0.1983\n",
      "Epoch 28/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9901 - loss: 0.0293 - val_accuracy: 0.9629 - val_loss: 0.1833\n",
      "Epoch 29/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9925 - loss: 0.0232 - val_accuracy: 0.9595 - val_loss: 0.1928\n",
      "Epoch 30/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9931 - loss: 0.0227 - val_accuracy: 0.9612 - val_loss: 0.1998\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.src.callbacks.history.History at 0x7af0a5a38940>"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Train the model\n",
    "model.fit(X_train, y_train, epochs=30, validation_data=(X_val, y_val))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "98e0df25",
   "metadata": {
    "papermill": {
     "duration": 0.091506,
     "end_time": "2024-09-23T20:32:18.843652",
     "exception": false,
     "start_time": "2024-09-23T20:32:18.752146",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "The model is performing much better on the training data than on the validation data, which means it's overfitting. High variance or overfitting happens when the model learns too much from the training data, including noise, and struggles to generalize to new data. To fix this, we will add a regularization term, which acts like a penalty and helps the model focus on the important patterns rather than memorizing the training data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "08e2603f",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-09-23T20:32:19.030765Z",
     "iopub.status.busy": "2024-09-23T20:32:19.030379Z",
     "iopub.status.idle": "2024-09-23T20:33:18.918322Z",
     "shell.execute_reply": "2024-09-23T20:33:18.917361Z"
    },
    "papermill": {
     "duration": 59.984798,
     "end_time": "2024-09-23T20:33:18.920407",
     "exception": false,
     "start_time": "2024-09-23T20:32:18.935609",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m6s\u001b[0m 3ms/step - accuracy: 0.7347 - loss: 0.9669 - val_accuracy: 0.9188 - val_loss: 0.3730\n",
      "Epoch 2/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9303 - loss: 0.3469 - val_accuracy: 0.9388 - val_loss: 0.3117\n",
      "Epoch 3/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9421 - loss: 0.3086 - val_accuracy: 0.9421 - val_loss: 0.3034\n",
      "Epoch 4/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9500 - loss: 0.2806 - val_accuracy: 0.9450 - val_loss: 0.2914\n",
      "Epoch 5/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9514 - loss: 0.2705 - val_accuracy: 0.9481 - val_loss: 0.2808\n",
      "Epoch 6/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9576 - loss: 0.2612 - val_accuracy: 0.9445 - val_loss: 0.2976\n",
      "Epoch 7/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9585 - loss: 0.2527 - val_accuracy: 0.9524 - val_loss: 0.2705\n",
      "Epoch 8/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 1ms/step - accuracy: 0.9617 - loss: 0.2497 - val_accuracy: 0.9479 - val_loss: 0.2814\n",
      "Epoch 9/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 1ms/step - accuracy: 0.9610 - loss: 0.2509 - val_accuracy: 0.9548 - val_loss: 0.2669\n",
      "Epoch 10/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 1ms/step - accuracy: 0.9629 - loss: 0.2452 - val_accuracy: 0.9540 - val_loss: 0.2668\n",
      "Epoch 11/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 1ms/step - accuracy: 0.9653 - loss: 0.2403 - val_accuracy: 0.9540 - val_loss: 0.2664\n",
      "Epoch 12/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9644 - loss: 0.2405 - val_accuracy: 0.9569 - val_loss: 0.2643\n",
      "Epoch 13/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9649 - loss: 0.2388 - val_accuracy: 0.9569 - val_loss: 0.2654\n",
      "Epoch 14/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9655 - loss: 0.2379 - val_accuracy: 0.9540 - val_loss: 0.2648\n",
      "Epoch 15/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 1ms/step - accuracy: 0.9685 - loss: 0.2305 - val_accuracy: 0.9598 - val_loss: 0.2581\n",
      "Epoch 16/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9685 - loss: 0.2313 - val_accuracy: 0.9562 - val_loss: 0.2596\n",
      "Epoch 17/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9672 - loss: 0.2328 - val_accuracy: 0.9588 - val_loss: 0.2589\n",
      "Epoch 18/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9671 - loss: 0.2310 - val_accuracy: 0.9583 - val_loss: 0.2591\n",
      "Epoch 19/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9685 - loss: 0.2316 - val_accuracy: 0.9588 - val_loss: 0.2618\n",
      "Epoch 20/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9684 - loss: 0.2293 - val_accuracy: 0.9552 - val_loss: 0.2666\n",
      "Epoch 21/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9696 - loss: 0.2302 - val_accuracy: 0.9595 - val_loss: 0.2536\n",
      "Epoch 22/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9698 - loss: 0.2277 - val_accuracy: 0.9579 - val_loss: 0.2574\n",
      "Epoch 23/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9695 - loss: 0.2265 - val_accuracy: 0.9576 - val_loss: 0.2559\n",
      "Epoch 24/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9698 - loss: 0.2248 - val_accuracy: 0.9600 - val_loss: 0.2511\n",
      "Epoch 25/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9709 - loss: 0.2235 - val_accuracy: 0.9576 - val_loss: 0.2541\n",
      "Epoch 26/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9698 - loss: 0.2252 - val_accuracy: 0.9595 - val_loss: 0.2576\n",
      "Epoch 27/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9713 - loss: 0.2243 - val_accuracy: 0.9610 - val_loss: 0.2571\n",
      "Epoch 28/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9695 - loss: 0.2269 - val_accuracy: 0.9555 - val_loss: 0.2646\n",
      "Epoch 29/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9678 - loss: 0.2312 - val_accuracy: 0.9540 - val_loss: 0.2690\n",
      "Epoch 30/30\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9717 - loss: 0.2242 - val_accuracy: 0.9600 - val_loss: 0.2556\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.src.callbacks.history.History at 0x7af0a40a1180>"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Define a model with regularization\n",
    "model = Sequential([\n",
    "    Dense(units=25, activation='relu', kernel_regularizer=L2(0.001)),\n",
    "    Dense(units=15, activation='relu', kernel_regularizer=L2(0.001)),\n",
    "    Dense(units=10, activation='softmax', kernel_regularizer=L2(0.001))\n",
    "])\n",
    "\n",
    "# Compile the model with an Adam optimizer and metrics\n",
    "model.compile(\n",
    "    optimizer = Adam(learning_rate=1e-3),\n",
    "    loss=SparseCategoricalCrossentropy(),\n",
    "    metrics=['accuracy']\n",
    ")\n",
    "\n",
    "# Train the model\n",
    "model.fit(X_train, y_train, epochs=30, validation_data=(X_val, y_val))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "57ae7043",
   "metadata": {
    "papermill": {
     "duration": 0.177149,
     "end_time": "2024-09-23T20:33:19.276169",
     "exception": false,
     "start_time": "2024-09-23T20:33:19.099020",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "The difference between training and validation performance has reduced, but there's still room for improvement in accuracy. To enhance the model, we'll increase its complexity while maintaining regularization and adding Dropout. This will help the model generalize better and improve overall performance."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "053d2167",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-09-23T20:33:19.630361Z",
     "iopub.status.busy": "2024-09-23T20:33:19.629584Z",
     "iopub.status.idle": "2024-09-23T20:35:48.193861Z",
     "shell.execute_reply": "2024-09-23T20:35:48.192912Z"
    },
    "papermill": {
     "duration": 148.743701,
     "end_time": "2024-09-23T20:35:48.195865",
     "exception": false,
     "start_time": "2024-09-23T20:33:19.452164",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m20s\u001b[0m 9ms/step - accuracy: 0.2249 - loss: 3.2612 - val_accuracy: 0.7579 - val_loss: 1.8216\n",
      "Epoch 2/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.6592 - loss: 1.9564 - val_accuracy: 0.8762 - val_loss: 1.3386\n",
      "Epoch 3/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.7966 - loss: 1.5292 - val_accuracy: 0.9102 - val_loss: 1.1505\n",
      "Epoch 4/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.8592 - loss: 1.3045 - val_accuracy: 0.9252 - val_loss: 1.0384\n",
      "Epoch 5/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.8905 - loss: 1.1631 - val_accuracy: 0.9393 - val_loss: 0.9573\n",
      "Epoch 6/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9096 - loss: 1.0526 - val_accuracy: 0.9474 - val_loss: 0.8932\n",
      "Epoch 7/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9195 - loss: 0.9859 - val_accuracy: 0.9510 - val_loss: 0.8361\n",
      "Epoch 8/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9313 - loss: 0.9080 - val_accuracy: 0.9543 - val_loss: 0.7878\n",
      "Epoch 9/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9364 - loss: 0.8561 - val_accuracy: 0.9588 - val_loss: 0.7434\n",
      "Epoch 10/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9440 - loss: 0.8008 - val_accuracy: 0.9605 - val_loss: 0.7025\n",
      "Epoch 11/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9457 - loss: 0.7533 - val_accuracy: 0.9631 - val_loss: 0.6665\n",
      "Epoch 12/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9524 - loss: 0.7056 - val_accuracy: 0.9667 - val_loss: 0.6285\n",
      "Epoch 13/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9540 - loss: 0.6781 - val_accuracy: 0.9664 - val_loss: 0.6041\n",
      "Epoch 14/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9562 - loss: 0.6453 - val_accuracy: 0.9693 - val_loss: 0.5725\n",
      "Epoch 15/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9602 - loss: 0.6082 - val_accuracy: 0.9690 - val_loss: 0.5477\n",
      "Epoch 16/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9592 - loss: 0.5846 - val_accuracy: 0.9700 - val_loss: 0.5251\n",
      "Epoch 17/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9648 - loss: 0.5541 - val_accuracy: 0.9710 - val_loss: 0.5034\n",
      "Epoch 18/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9646 - loss: 0.5286 - val_accuracy: 0.9743 - val_loss: 0.4805\n",
      "Epoch 19/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9684 - loss: 0.5007 - val_accuracy: 0.9733 - val_loss: 0.4658\n",
      "Epoch 20/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9667 - loss: 0.4913 - val_accuracy: 0.9726 - val_loss: 0.4525\n",
      "Epoch 21/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9696 - loss: 0.4690 - val_accuracy: 0.9752 - val_loss: 0.4326\n",
      "Epoch 22/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9690 - loss: 0.4552 - val_accuracy: 0.9738 - val_loss: 0.4246\n",
      "Epoch 23/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 2ms/step - accuracy: 0.9722 - loss: 0.4340 - val_accuracy: 0.9738 - val_loss: 0.4103\n",
      "Epoch 24/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9722 - loss: 0.4254 - val_accuracy: 0.9743 - val_loss: 0.4012\n",
      "Epoch 25/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9735 - loss: 0.4070 - val_accuracy: 0.9755 - val_loss: 0.3863\n",
      "Epoch 26/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9748 - loss: 0.3946 - val_accuracy: 0.9771 - val_loss: 0.3760\n",
      "Epoch 27/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9744 - loss: 0.3847 - val_accuracy: 0.9760 - val_loss: 0.3694\n",
      "Epoch 28/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9756 - loss: 0.3712 - val_accuracy: 0.9767 - val_loss: 0.3597\n",
      "Epoch 29/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9776 - loss: 0.3597 - val_accuracy: 0.9762 - val_loss: 0.3520\n",
      "Epoch 30/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9777 - loss: 0.3510 - val_accuracy: 0.9767 - val_loss: 0.3445\n",
      "Epoch 31/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9778 - loss: 0.3415 - val_accuracy: 0.9743 - val_loss: 0.3361\n",
      "Epoch 32/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9789 - loss: 0.3326 - val_accuracy: 0.9786 - val_loss: 0.3321\n",
      "Epoch 33/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9794 - loss: 0.3276 - val_accuracy: 0.9786 - val_loss: 0.3246\n",
      "Epoch 34/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9804 - loss: 0.3175 - val_accuracy: 0.9764 - val_loss: 0.3200\n",
      "Epoch 35/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9804 - loss: 0.3129 - val_accuracy: 0.9764 - val_loss: 0.3130\n",
      "Epoch 36/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9804 - loss: 0.3044 - val_accuracy: 0.9764 - val_loss: 0.3153\n",
      "Epoch 37/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9832 - loss: 0.2980 - val_accuracy: 0.9769 - val_loss: 0.3052\n",
      "Epoch 38/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9801 - loss: 0.2958 - val_accuracy: 0.9767 - val_loss: 0.2995\n",
      "Epoch 39/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9817 - loss: 0.2904 - val_accuracy: 0.9774 - val_loss: 0.2969\n",
      "Epoch 40/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9828 - loss: 0.2841 - val_accuracy: 0.9774 - val_loss: 0.2942\n",
      "Epoch 41/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9834 - loss: 0.2810 - val_accuracy: 0.9781 - val_loss: 0.2924\n",
      "Epoch 42/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9822 - loss: 0.2798 - val_accuracy: 0.9752 - val_loss: 0.2897\n",
      "Epoch 43/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9836 - loss: 0.2732 - val_accuracy: 0.9781 - val_loss: 0.2860\n",
      "Epoch 44/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9849 - loss: 0.2682 - val_accuracy: 0.9767 - val_loss: 0.2822\n",
      "Epoch 45/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9846 - loss: 0.2635 - val_accuracy: 0.9771 - val_loss: 0.2773\n",
      "Epoch 46/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9835 - loss: 0.2627 - val_accuracy: 0.9764 - val_loss: 0.2768\n",
      "Epoch 47/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9845 - loss: 0.2591 - val_accuracy: 0.9781 - val_loss: 0.2732\n",
      "Epoch 48/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9838 - loss: 0.2595 - val_accuracy: 0.9793 - val_loss: 0.2715\n",
      "Epoch 49/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9850 - loss: 0.2541 - val_accuracy: 0.9795 - val_loss: 0.2662\n",
      "Epoch 50/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9859 - loss: 0.2497 - val_accuracy: 0.9781 - val_loss: 0.2659\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.src.callbacks.history.History at 0x7af03c5bf430>"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Define a deeper and more complex model\n",
    "model = Sequential([\n",
    "    Dense(units=512, activation='relu', kernel_regularizer=L2(0.001)),  # Increased units\n",
    "    Dropout(0.4),  # Add dropout to prevent overfitting\n",
    "    Dense(units=256, activation='relu', kernel_regularizer=L2(0.001)),\n",
    "    Dropout(0.4),\n",
    "    Dense(units=128, activation='relu', kernel_regularizer=L2(0.001)),\n",
    "    Dropout(0.3),\n",
    "    Dense(units=64, activation='relu', kernel_regularizer=L2(0.001)),\n",
    "    Dropout(0.3),\n",
    "    Dense(units=32, activation='relu', kernel_regularizer=L2(0.001)),\n",
    "    Dense(units=10, activation='softmax', kernel_regularizer=L2(0.001))  \n",
    "])\n",
    "\n",
    "# Compile the model with an Adam optimizer and metrics\n",
    "model.compile(\n",
    "    optimizer=Adam(learning_rate=5e-5),  # Reduced learning rate for better convergence\n",
    "    loss=SparseCategoricalCrossentropy(),\n",
    "    metrics=['accuracy']\n",
    ")\n",
    "\n",
    "# Train the model\n",
    "model.fit(X_train, y_train, epochs=50, validation_data=(X_val, y_val))  # Increased epochs"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "12f9e4e1",
   "metadata": {
    "papermill": {
     "duration": 0.380334,
     "end_time": "2024-09-23T20:35:48.951897",
     "exception": false,
     "start_time": "2024-09-23T20:35:48.571563",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "The results are promising, but now let's take it a step further by trying a convolutional neural network (CNN)."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c90bc963",
   "metadata": {
    "papermill": {
     "duration": 0.381644,
     "end_time": "2024-09-23T20:35:49.749258",
     "exception": false,
     "start_time": "2024-09-23T20:35:49.367614",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "<h2 style=\"text-align: center; \n",
    "           border-radius: 12px; \n",
    "           padding: 15px; \n",
    "           margin: 0; \n",
    "           font-size: 120%; \n",
    "           font-family: Arial, sans-serif; \n",
    "           background-color: #046B85; \n",
    "           color: #F5F5F5; \n",
    "           box-shadow: 0 3px 6px rgba(0, 0, 0, 0.2); \n",
    "           border: 1px solid #034C63;\">\n",
    "   <b>5. Convolutional Neural Network</b>\n",
    "</h2>"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cbe88e71",
   "metadata": {
    "papermill": {
     "duration": 0.373731,
     "end_time": "2024-09-23T20:35:50.511122",
     "exception": false,
     "start_time": "2024-09-23T20:35:50.137391",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "Convolutional Neural Networks (CNNs) are powerful for image recognition tasks because they can capture spatial features, like edges, patterns, and shapes, making them much more effective than dense neural networks for working with images."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "81b67efd",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-09-23T20:35:51.259653Z",
     "iopub.status.busy": "2024-09-23T20:35:51.259247Z",
     "iopub.status.idle": "2024-09-23T20:35:51.264692Z",
     "shell.execute_reply": "2024-09-23T20:35:51.263817Z"
    },
    "papermill": {
     "duration": 0.379764,
     "end_time": "2024-09-23T20:35:51.266732",
     "exception": false,
     "start_time": "2024-09-23T20:35:50.886968",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Convert DataFrames to NumPy arrays\n",
    "X_train = X_train.to_numpy().reshape(-1, 28, 28, 1)\n",
    "X_val = X_val.to_numpy().reshape(-1, 28, 28, 1)\n",
    "X_test = X_test.to_numpy().reshape(-1, 28, 28, 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "ff70b031",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-09-23T20:35:52.060403Z",
     "iopub.status.busy": "2024-09-23T20:35:52.060026Z",
     "iopub.status.idle": "2024-09-23T20:38:37.089730Z",
     "shell.execute_reply": "2024-09-23T20:38:37.088802Z"
    },
    "papermill": {
     "duration": 165.411027,
     "end_time": "2024-09-23T20:38:37.091631",
     "exception": false,
     "start_time": "2024-09-23T20:35:51.680604",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m20s\u001b[0m 9ms/step - accuracy: 0.4781 - loss: 1.5454 - val_accuracy: 0.9445 - val_loss: 0.2335\n",
      "Epoch 2/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 3ms/step - accuracy: 0.9020 - loss: 0.3391 - val_accuracy: 0.9619 - val_loss: 0.1565\n",
      "Epoch 3/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9350 - loss: 0.2375 - val_accuracy: 0.9705 - val_loss: 0.1283\n",
      "Epoch 4/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9491 - loss: 0.1907 - val_accuracy: 0.9767 - val_loss: 0.1094\n",
      "Epoch 5/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9571 - loss: 0.1682 - val_accuracy: 0.9783 - val_loss: 0.0992\n",
      "Epoch 6/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9639 - loss: 0.1463 - val_accuracy: 0.9812 - val_loss: 0.0934\n",
      "Epoch 7/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9680 - loss: 0.1304 - val_accuracy: 0.9810 - val_loss: 0.0877\n",
      "Epoch 8/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9713 - loss: 0.1222 - val_accuracy: 0.9840 - val_loss: 0.0813\n",
      "Epoch 9/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9739 - loss: 0.1153 - val_accuracy: 0.9843 - val_loss: 0.0799\n",
      "Epoch 10/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9741 - loss: 0.1108 - val_accuracy: 0.9857 - val_loss: 0.0766\n",
      "Epoch 11/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 3ms/step - accuracy: 0.9758 - loss: 0.1066 - val_accuracy: 0.9864 - val_loss: 0.0730\n",
      "Epoch 12/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9786 - loss: 0.0988 - val_accuracy: 0.9869 - val_loss: 0.0729\n",
      "Epoch 13/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9803 - loss: 0.0939 - val_accuracy: 0.9881 - val_loss: 0.0691\n",
      "Epoch 14/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9816 - loss: 0.0893 - val_accuracy: 0.9876 - val_loss: 0.0699\n",
      "Epoch 15/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 3ms/step - accuracy: 0.9839 - loss: 0.0846 - val_accuracy: 0.9879 - val_loss: 0.0684\n",
      "Epoch 16/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 3ms/step - accuracy: 0.9828 - loss: 0.0865 - val_accuracy: 0.9893 - val_loss: 0.0654\n",
      "Epoch 17/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 3ms/step - accuracy: 0.9850 - loss: 0.0809 - val_accuracy: 0.9888 - val_loss: 0.0661\n",
      "Epoch 18/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9843 - loss: 0.0806 - val_accuracy: 0.9895 - val_loss: 0.0644\n",
      "Epoch 19/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9848 - loss: 0.0788 - val_accuracy: 0.9893 - val_loss: 0.0633\n",
      "Epoch 20/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 3ms/step - accuracy: 0.9848 - loss: 0.0798 - val_accuracy: 0.9900 - val_loss: 0.0613\n",
      "Epoch 21/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 3ms/step - accuracy: 0.9866 - loss: 0.0734 - val_accuracy: 0.9890 - val_loss: 0.0632\n",
      "Epoch 22/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 3ms/step - accuracy: 0.9862 - loss: 0.0719 - val_accuracy: 0.9902 - val_loss: 0.0603\n",
      "Epoch 23/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 3ms/step - accuracy: 0.9876 - loss: 0.0710 - val_accuracy: 0.9900 - val_loss: 0.0610\n",
      "Epoch 24/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 3ms/step - accuracy: 0.9857 - loss: 0.0738 - val_accuracy: 0.9917 - val_loss: 0.0581\n",
      "Epoch 25/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 3ms/step - accuracy: 0.9874 - loss: 0.0713 - val_accuracy: 0.9910 - val_loss: 0.0583\n",
      "Epoch 26/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9881 - loss: 0.0645 - val_accuracy: 0.9902 - val_loss: 0.0587\n",
      "Epoch 27/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9885 - loss: 0.0662 - val_accuracy: 0.9914 - val_loss: 0.0591\n",
      "Epoch 28/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 3ms/step - accuracy: 0.9877 - loss: 0.0668 - val_accuracy: 0.9919 - val_loss: 0.0548\n",
      "Epoch 29/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9899 - loss: 0.0625 - val_accuracy: 0.9912 - val_loss: 0.0557\n",
      "Epoch 30/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 3ms/step - accuracy: 0.9905 - loss: 0.0617 - val_accuracy: 0.9893 - val_loss: 0.0613\n",
      "Epoch 31/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 3ms/step - accuracy: 0.9889 - loss: 0.0626 - val_accuracy: 0.9910 - val_loss: 0.0563\n",
      "Epoch 32/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 3ms/step - accuracy: 0.9897 - loss: 0.0634 - val_accuracy: 0.9910 - val_loss: 0.0560\n",
      "Epoch 33/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 3ms/step - accuracy: 0.9899 - loss: 0.0619 - val_accuracy: 0.9929 - val_loss: 0.0549\n",
      "Epoch 34/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9907 - loss: 0.0597 - val_accuracy: 0.9912 - val_loss: 0.0544\n",
      "Epoch 35/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9899 - loss: 0.0588 - val_accuracy: 0.9919 - val_loss: 0.0553\n",
      "Epoch 36/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 3ms/step - accuracy: 0.9894 - loss: 0.0599 - val_accuracy: 0.9917 - val_loss: 0.0539\n",
      "Epoch 37/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9919 - loss: 0.0538 - val_accuracy: 0.9895 - val_loss: 0.0554\n",
      "Epoch 38/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9912 - loss: 0.0552 - val_accuracy: 0.9919 - val_loss: 0.0529\n",
      "Epoch 39/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9913 - loss: 0.0537 - val_accuracy: 0.9931 - val_loss: 0.0529\n",
      "Epoch 40/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9917 - loss: 0.0551 - val_accuracy: 0.9919 - val_loss: 0.0551\n",
      "Epoch 41/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9914 - loss: 0.0551 - val_accuracy: 0.9910 - val_loss: 0.0545\n",
      "Epoch 42/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9905 - loss: 0.0548 - val_accuracy: 0.9924 - val_loss: 0.0510\n",
      "Epoch 43/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 3ms/step - accuracy: 0.9928 - loss: 0.0514 - val_accuracy: 0.9919 - val_loss: 0.0539\n",
      "Epoch 44/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9925 - loss: 0.0496 - val_accuracy: 0.9919 - val_loss: 0.0532\n",
      "Epoch 45/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9916 - loss: 0.0524 - val_accuracy: 0.9926 - val_loss: 0.0512\n",
      "Epoch 46/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9923 - loss: 0.0502 - val_accuracy: 0.9931 - val_loss: 0.0498\n",
      "Epoch 47/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9923 - loss: 0.0511 - val_accuracy: 0.9921 - val_loss: 0.0497\n",
      "Epoch 48/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9916 - loss: 0.0516 - val_accuracy: 0.9910 - val_loss: 0.0525\n",
      "Epoch 49/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9921 - loss: 0.0494 - val_accuracy: 0.9933 - val_loss: 0.0472\n",
      "Epoch 50/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9924 - loss: 0.0496 - val_accuracy: 0.9931 - val_loss: 0.0483\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.src.callbacks.history.History at 0x7af0306de1a0>"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Define a model with Conv2D layers\n",
    "model = Sequential([\n",
    "    Conv2D(32, kernel_size=(3, 3), activation='relu', padding = 'Same', kernel_regularizer=L2(0.0001), input_shape=(28, 28, 1)),    # First conv layer\n",
    "    MaxPooling2D(pool_size=(2, 2)),   # Pooling layer\n",
    "    Dropout(0.2),  # Dropout to reduce overfitting\n",
    "    \n",
    "    Conv2D(64, kernel_size=(3, 3), activation='relu', padding = 'Same', kernel_regularizer=L2(0.0001)),  # Second conv layer\n",
    "    MaxPooling2D(pool_size=(2, 2)),\n",
    "    Dropout(0.3),\n",
    "    \n",
    "    Conv2D(128, kernel_size=(3, 3), activation='relu', padding = 'Same', kernel_regularizer=L2(0.0001)),   # Third conv layer\n",
    "    MaxPooling2D(pool_size=(2, 2)),\n",
    "    Dropout(0.3),\n",
    "    \n",
    "    Flatten(),    # Flatten the output for dense layers\n",
    "    Dense(64, activation='relu', kernel_regularizer=L2(0.0001)),  # Fully connected layer\n",
    "    Dropout(0.2),\n",
    "    Dense(10, activation='softmax')  # Output layer\n",
    "])\n",
    "\n",
    "# Compile the model\n",
    "model.compile(\n",
    "    optimizer=Adam(learning_rate=1e-4),   \n",
    "    loss=SparseCategoricalCrossentropy(),\n",
    "    metrics=['accuracy']\n",
    ")\n",
    "\n",
    "# Train the model\n",
    "model.fit(X_train, y_train, epochs=50, validation_data=(X_val, y_val))  "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f86170f4",
   "metadata": {
    "papermill": {
     "duration": 0.634824,
     "end_time": "2024-09-23T20:38:38.324181",
     "exception": false,
     "start_time": "2024-09-23T20:38:37.689357",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "### Data Augmentation"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "55313f20",
   "metadata": {
    "papermill": {
     "duration": 0.593661,
     "end_time": "2024-09-23T20:38:39.514000",
     "exception": false,
     "start_time": "2024-09-23T20:38:38.920339",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "Data augmentation is a powerful technique used to improve the performance and generalization of machine learning models, especially in image recognition tasks like digit recognition. The idea behind data augmentation is to artificially increase the size and diversity of the training dataset by applying random transformations to the original images, such as rotations, zooms, shifts, or flips. This helps the model learn more robust features and reduces the risk of overfitting.\n",
    "\n",
    "Since our training dataset may not be large or varied enough, data augmentation allows us to create new variations of the existing images, simulating real-world scenarios where digits might be slightly tilted, resized, or shifted. This technique makes the model more adaptable and better at recognizing digits in various conditions.\n",
    "\n",
    "For this project, we use data augmentation to randomly rotate, shift, or zoom in on images, enhancing the model's ability to generalize well to unseen data. This ensures that the model doesn't simply memorize the training images but learns patterns that can be applied to new data."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "351ff33f",
   "metadata": {
    "papermill": {
     "duration": 0.595991,
     "end_time": "2024-09-23T20:38:40.742877",
     "exception": false,
     "start_time": "2024-09-23T20:38:40.146886",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "#### Key Aspects:\n",
    "- **Rotation Range (5 degrees)**: Digit recognition can benefit from slight rotations as handwritten digits may not be perfectly aligned. A small range like 5 degrees keeps the transformation realistic without distorting the digits.\n",
    "\n",
    "- **Zoom Range (0.1)**: A 10% zoom allows the model to learn from digits that may be written smaller or larger, which adds robustness.\n",
    "\n",
    "- **Width/Height Shift (0.1)**: Small translations mimic slight variations in where digits might appear in images. This helps the model handle digits that are not perfectly centered."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "f6b09463",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-09-23T20:38:41.933365Z",
     "iopub.status.busy": "2024-09-23T20:38:41.932513Z",
     "iopub.status.idle": "2024-09-23T20:38:42.023318Z",
     "shell.execute_reply": "2024-09-23T20:38:42.022490Z"
    },
    "papermill": {
     "duration": 0.689323,
     "end_time": "2024-09-23T20:38:42.025664",
     "exception": false,
     "start_time": "2024-09-23T20:38:41.336341",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "datagen = ImageDataGenerator(\n",
    "        rotation_range=5,         # Randomly rotate images by up to 5 degrees\n",
    "        zoom_range=0.1,            # Randomly zoom images by up to 10%\n",
    "        width_shift_range=0.1,     # Randomly shift images horizontally by up to 10% of width\n",
    "        height_shift_range=0.1,    # Randomly shift images vertically by up to 10% of height\n",
    ")\n",
    "\n",
    "datagen.fit(X_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "daf09777",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-09-23T20:38:43.265013Z",
     "iopub.status.busy": "2024-09-23T20:38:43.263883Z",
     "iopub.status.idle": "2024-09-23T20:52:13.517466Z",
     "shell.execute_reply": "2024-09-23T20:52:13.516414Z"
    },
    "papermill": {
     "duration": 810.857892,
     "end_time": "2024-09-23T20:52:13.520234",
     "exception": false,
     "start_time": "2024-09-23T20:38:42.662342",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m37s\u001b[0m 21ms/step - accuracy: 0.2239 - loss: 2.4818 - val_accuracy: 0.8805 - val_loss: 0.5993\n",
      "Epoch 2/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 13ms/step - accuracy: 0.6583 - loss: 1.0803 - val_accuracy: 0.9671 - val_loss: 0.1728\n",
      "Epoch 3/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 13ms/step - accuracy: 0.8422 - loss: 0.5669 - val_accuracy: 0.9798 - val_loss: 0.1265\n",
      "Epoch 4/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m21s\u001b[0m 13ms/step - accuracy: 0.8963 - loss: 0.4185 - val_accuracy: 0.9805 - val_loss: 0.1196\n",
      "Epoch 5/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m16s\u001b[0m 13ms/step - accuracy: 0.9281 - loss: 0.3166 - val_accuracy: 0.9860 - val_loss: 0.1067\n",
      "Epoch 6/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 13ms/step - accuracy: 0.9379 - loss: 0.2841 - val_accuracy: 0.9864 - val_loss: 0.1080\n",
      "Epoch 7/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m16s\u001b[0m 13ms/step - accuracy: 0.9507 - loss: 0.2430 - val_accuracy: 0.9867 - val_loss: 0.1063\n",
      "Epoch 8/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 13ms/step - accuracy: 0.9538 - loss: 0.2288 - val_accuracy: 0.9890 - val_loss: 0.0964\n",
      "Epoch 9/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m16s\u001b[0m 14ms/step - accuracy: 0.9591 - loss: 0.2085 - val_accuracy: 0.9890 - val_loss: 0.0937\n",
      "Epoch 10/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 13ms/step - accuracy: 0.9640 - loss: 0.1889 - val_accuracy: 0.9862 - val_loss: 0.1053\n",
      "Epoch 11/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m16s\u001b[0m 13ms/step - accuracy: 0.9660 - loss: 0.1884 - val_accuracy: 0.9895 - val_loss: 0.0859\n",
      "Epoch 12/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m16s\u001b[0m 13ms/step - accuracy: 0.9702 - loss: 0.1695 - val_accuracy: 0.9905 - val_loss: 0.0879\n",
      "Epoch 13/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m16s\u001b[0m 14ms/step - accuracy: 0.9734 - loss: 0.1613 - val_accuracy: 0.9902 - val_loss: 0.0887\n",
      "Epoch 14/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m16s\u001b[0m 13ms/step - accuracy: 0.9752 - loss: 0.1517 - val_accuracy: 0.9907 - val_loss: 0.0839\n",
      "Epoch 15/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m16s\u001b[0m 13ms/step - accuracy: 0.9737 - loss: 0.1550 - val_accuracy: 0.9910 - val_loss: 0.0843\n",
      "Epoch 16/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 13ms/step - accuracy: 0.9763 - loss: 0.1453 - val_accuracy: 0.9910 - val_loss: 0.0834\n",
      "Epoch 17/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 13ms/step - accuracy: 0.9780 - loss: 0.1408 - val_accuracy: 0.9910 - val_loss: 0.0853\n",
      "Epoch 18/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 13ms/step - accuracy: 0.9782 - loss: 0.1396 - val_accuracy: 0.9914 - val_loss: 0.0817\n",
      "Epoch 19/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m16s\u001b[0m 13ms/step - accuracy: 0.9800 - loss: 0.1321 - val_accuracy: 0.9936 - val_loss: 0.0785\n",
      "Epoch 20/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 13ms/step - accuracy: 0.9789 - loss: 0.1339 - val_accuracy: 0.9924 - val_loss: 0.0819\n",
      "Epoch 21/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m16s\u001b[0m 14ms/step - accuracy: 0.9808 - loss: 0.1239 - val_accuracy: 0.9933 - val_loss: 0.0806\n",
      "Epoch 22/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m16s\u001b[0m 13ms/step - accuracy: 0.9815 - loss: 0.1226 - val_accuracy: 0.9921 - val_loss: 0.0834\n",
      "Epoch 23/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m16s\u001b[0m 13ms/step - accuracy: 0.9822 - loss: 0.1268 - val_accuracy: 0.9924 - val_loss: 0.0807\n",
      "Epoch 24/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m16s\u001b[0m 13ms/step - accuracy: 0.9794 - loss: 0.1265 - val_accuracy: 0.9929 - val_loss: 0.0757\n",
      "Epoch 25/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m16s\u001b[0m 13ms/step - accuracy: 0.9841 - loss: 0.1145 - val_accuracy: 0.9924 - val_loss: 0.0771\n",
      "Epoch 26/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m16s\u001b[0m 13ms/step - accuracy: 0.9821 - loss: 0.1197 - val_accuracy: 0.9936 - val_loss: 0.0768\n",
      "Epoch 27/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m16s\u001b[0m 13ms/step - accuracy: 0.9837 - loss: 0.1139 - val_accuracy: 0.9919 - val_loss: 0.0774\n",
      "Epoch 28/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 13ms/step - accuracy: 0.9853 - loss: 0.1067 - val_accuracy: 0.9929 - val_loss: 0.0748\n",
      "Epoch 29/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m16s\u001b[0m 13ms/step - accuracy: 0.9851 - loss: 0.1069 - val_accuracy: 0.9936 - val_loss: 0.0756\n",
      "Epoch 30/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m16s\u001b[0m 13ms/step - accuracy: 0.9853 - loss: 0.1097 - val_accuracy: 0.9936 - val_loss: 0.0752\n",
      "Epoch 31/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m20s\u001b[0m 13ms/step - accuracy: 0.9857 - loss: 0.1054 - val_accuracy: 0.9929 - val_loss: 0.0758\n",
      "Epoch 32/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 13ms/step - accuracy: 0.9856 - loss: 0.1040 - val_accuracy: 0.9931 - val_loss: 0.0735\n",
      "Epoch 33/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m16s\u001b[0m 13ms/step - accuracy: 0.9859 - loss: 0.1042 - val_accuracy: 0.9936 - val_loss: 0.0734\n",
      "Epoch 34/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 13ms/step - accuracy: 0.9848 - loss: 0.1074 - val_accuracy: 0.9948 - val_loss: 0.0737\n",
      "Epoch 35/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m16s\u001b[0m 13ms/step - accuracy: 0.9871 - loss: 0.0995 - val_accuracy: 0.9933 - val_loss: 0.0732\n",
      "Epoch 36/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 13ms/step - accuracy: 0.9865 - loss: 0.1009 - val_accuracy: 0.9940 - val_loss: 0.0744\n",
      "Epoch 37/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m16s\u001b[0m 13ms/step - accuracy: 0.9872 - loss: 0.0983 - val_accuracy: 0.9933 - val_loss: 0.0723\n",
      "Epoch 38/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m16s\u001b[0m 13ms/step - accuracy: 0.9866 - loss: 0.0979 - val_accuracy: 0.9936 - val_loss: 0.0736\n",
      "Epoch 39/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m16s\u001b[0m 13ms/step - accuracy: 0.9871 - loss: 0.1002 - val_accuracy: 0.9933 - val_loss: 0.0744\n",
      "Epoch 40/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 13ms/step - accuracy: 0.9875 - loss: 0.0992 - val_accuracy: 0.9940 - val_loss: 0.0733\n",
      "Epoch 41/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 13ms/step - accuracy: 0.9883 - loss: 0.0921 - val_accuracy: 0.9945 - val_loss: 0.0706\n",
      "Epoch 42/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 13ms/step - accuracy: 0.9882 - loss: 0.0920 - val_accuracy: 0.9929 - val_loss: 0.0723\n",
      "Epoch 43/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 13ms/step - accuracy: 0.9880 - loss: 0.0938 - val_accuracy: 0.9933 - val_loss: 0.0726\n",
      "Epoch 44/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 13ms/step - accuracy: 0.9885 - loss: 0.0929 - val_accuracy: 0.9943 - val_loss: 0.0712\n",
      "Epoch 45/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 13ms/step - accuracy: 0.9883 - loss: 0.0951 - val_accuracy: 0.9945 - val_loss: 0.0706\n",
      "Epoch 46/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 13ms/step - accuracy: 0.9886 - loss: 0.0944 - val_accuracy: 0.9926 - val_loss: 0.0740\n",
      "Epoch 47/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 13ms/step - accuracy: 0.9882 - loss: 0.0924 - val_accuracy: 0.9940 - val_loss: 0.0709\n",
      "Epoch 48/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m16s\u001b[0m 13ms/step - accuracy: 0.9886 - loss: 0.0904 - val_accuracy: 0.9943 - val_loss: 0.0706\n",
      "Epoch 49/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 13ms/step - accuracy: 0.9900 - loss: 0.0847 - val_accuracy: 0.9940 - val_loss: 0.0673\n",
      "Epoch 50/50\n",
      "\u001b[1m1182/1182\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 13ms/step - accuracy: 0.9882 - loss: 0.0917 - val_accuracy: 0.9940 - val_loss: 0.0708\n"
     ]
    }
   ],
   "source": [
    "# Define a more complex model with adjusted architecture\n",
    "model = Sequential([\n",
    "    # First convolutional block\n",
    "    Conv2D(64, kernel_size=(3, 3), activation='relu', padding = 'Same', kernel_regularizer=L2(0.0001), input_shape=(28, 28, 1)),\n",
    "    BatchNormalization(),\n",
    "    MaxPooling2D(pool_size=(2, 2)),\n",
    "    Dropout(0.3),\n",
    "    \n",
    "    # Second convolutional block\n",
    "    Conv2D(128, kernel_size=(3, 3), activation='relu', padding = 'Same', kernel_regularizer=L2(0.0001)),\n",
    "    BatchNormalization(),\n",
    "    MaxPooling2D(pool_size=(2, 2)),\n",
    "    Dropout(0.3),\n",
    "    \n",
    "    # Third convolutional block\n",
    "    Conv2D(256, kernel_size=(3, 3), activation='relu', padding = 'Same', kernel_regularizer=L2(0.0001)),\n",
    "    BatchNormalization(),\n",
    "    MaxPooling2D(pool_size=(2, 2)),\n",
    "    Dropout(0.4),\n",
    "    \n",
    "    # Flatten and fully connected layers\n",
    "    Flatten(),\n",
    "    Dense(128, activation='relu', kernel_regularizer=L2(0.0001)),\n",
    "    Dropout(0.4),\n",
    "    Dense(64, activation='relu', kernel_regularizer=L2(0.0001)),\n",
    "    Dropout(0.3),\n",
    "    \n",
    "    # Output layer for classification\n",
    "    Dense(10, activation='softmax')\n",
    "])\n",
    "\n",
    "# Compile the model\n",
    "model.compile(\n",
    "    optimizer=Adam(learning_rate=1e-4),\n",
    "    loss=SparseCategoricalCrossentropy(),\n",
    "    metrics=['accuracy']\n",
    ")\n",
    "\n",
    "# Train the model\n",
    "history = model.fit(datagen.flow(X_train, y_train, batch_size=32),\n",
    "                    epochs=50,   \n",
    "                    validation_data=(X_val, y_val))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "41a14a48",
   "metadata": {
    "papermill": {
     "duration": 1.830883,
     "end_time": "2024-09-23T20:52:17.141189",
     "exception": false,
     "start_time": "2024-09-23T20:52:15.310306",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "### Model Evaluation"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8ffc5bfe",
   "metadata": {
    "papermill": {
     "duration": 1.792479,
     "end_time": "2024-09-23T20:52:20.720902",
     "exception": false,
     "start_time": "2024-09-23T20:52:18.928423",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "#### Training and Validation Curves"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "a52f0618",
   "metadata": {
    "_kg_hide-input": true,
    "execution": {
     "iopub.execute_input": "2024-09-23T20:52:24.273883Z",
     "iopub.status.busy": "2024-09-23T20:52:24.272969Z",
     "iopub.status.idle": "2024-09-23T20:52:24.781355Z",
     "shell.execute_reply": "2024-09-23T20:52:24.780439Z"
    },
    "papermill": {
     "duration": 2.319513,
     "end_time": "2024-09-23T20:52:24.783461",
     "exception": false,
     "start_time": "2024-09-23T20:52:22.463948",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot accuracy\n",
    "plt.plot(history.history['accuracy'], label='train accuracy')\n",
    "plt.plot(history.history['val_accuracy'], label='validation accuracy')\n",
    "plt.legend()\n",
    "plt.show()\n",
    "\n",
    "# Plot loss\n",
    "plt.plot(history.history['loss'], label='train loss')\n",
    "plt.plot(history.history['val_loss'], label='validation loss')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8e5eb8d6",
   "metadata": {
    "papermill": {
     "duration": 1.810507,
     "end_time": "2024-09-23T20:52:28.380724",
     "exception": false,
     "start_time": "2024-09-23T20:52:26.570217",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "#### Confusion Matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "369d1d49",
   "metadata": {
    "_kg_hide-input": true,
    "execution": {
     "iopub.execute_input": "2024-09-23T20:52:31.967765Z",
     "iopub.status.busy": "2024-09-23T20:52:31.966840Z",
     "iopub.status.idle": "2024-09-23T20:52:33.536788Z",
     "shell.execute_reply": "2024-09-23T20:52:33.535891Z"
    },
    "papermill": {
     "duration": 3.365549,
     "end_time": "2024-09-23T20:52:33.539007",
     "exception": false,
     "start_time": "2024-09-23T20:52:30.173458",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[1m132/132\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 4ms/step\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 800x600 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# 1. Make predictions on the validation or test set\n",
    "y_pred_probs = model.predict(X_val)  # Predict probabilities\n",
    "y_pred = np.argmax(y_pred_probs, axis=1)  # Convert probabilities to class labels\n",
    "\n",
    "# 2. Generate the confusion matrix\n",
    "conf_matrix = confusion_matrix(y_val, y_pred)\n",
    "\n",
    "# 3. Define class names (digits 0 to 9)\n",
    "class_names = [i for i in range(10)]\n",
    "\n",
    "# 4. Plot the confusion matrix\n",
    "plt.figure(figsize=(8, 6))\n",
    "sns.heatmap(conf_matrix, annot=True, fmt='d', cmap='Blues', xticklabels=class_names, yticklabels=class_names)\n",
    "plt.xlabel('Predicted Label')\n",
    "plt.ylabel('True Label')\n",
    "plt.title('Confusion Matrix')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7321fbb5",
   "metadata": {
    "papermill": {
     "duration": 1.836049,
     "end_time": "2024-09-23T20:52:37.165452",
     "exception": false,
     "start_time": "2024-09-23T20:52:35.329403",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "### Submitting Predictions to the Kaggle 'Digit Recognizer' Competition"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "12474b23",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-09-23T20:52:40.763718Z",
     "iopub.status.busy": "2024-09-23T20:52:40.763353Z",
     "iopub.status.idle": "2024-09-23T20:52:42.823310Z",
     "shell.execute_reply": "2024-09-23T20:52:42.822465Z"
    },
    "papermill": {
     "duration": 3.859666,
     "end_time": "2024-09-23T20:52:42.825361",
     "exception": false,
     "start_time": "2024-09-23T20:52:38.965695",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[1m219/219\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 5ms/step\n"
     ]
    }
   ],
   "source": [
    "predictions = model.predict(X_test, batch_size=128)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "5d871a82",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-09-23T20:52:46.369255Z",
     "iopub.status.busy": "2024-09-23T20:52:46.368882Z",
     "iopub.status.idle": "2024-09-23T20:52:46.373845Z",
     "shell.execute_reply": "2024-09-23T20:52:46.373014Z"
    },
    "papermill": {
     "duration": 1.812854,
     "end_time": "2024-09-23T20:52:46.375747",
     "exception": false,
     "start_time": "2024-09-23T20:52:44.562893",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "predicted_classes = np.argmax(predictions, axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "e4be637f",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2024-09-23T20:52:50.062704Z",
     "iopub.status.busy": "2024-09-23T20:52:50.062343Z",
     "iopub.status.idle": "2024-09-23T20:52:50.102877Z",
     "shell.execute_reply": "2024-09-23T20:52:50.102148Z"
    },
    "papermill": {
     "duration": 1.863097,
     "end_time": "2024-09-23T20:52:50.104830",
     "exception": false,
     "start_time": "2024-09-23T20:52:48.241733",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "submission_df = pd.DataFrame({\n",
    "    'ImageId': range(1, len(predicted_classes) + 1),\n",
    "    'Label': predicted_classes\n",
    "})\n",
    "submission_df.to_csv('submission.csv', index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3c7a4549",
   "metadata": {},
   "source": [
    "**The submission achieved an impressive accuracy score of 0.99514 on the Kaggle leaderboard.**"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1aca738d",
   "metadata": {},
   "source": [
    "<div style=\"border: 2px solid #84172D; \n",
    "            padding: 20px; \n",
    "            border-radius: 12px; \n",
    "            background-color: #f4f4f4; \n",
    "            color: #333; \n",
    "            box-shadow: 0 4px 8px rgba(0, 0, 0, 0.1);\">\n",
    "    <p style=\"font-size: 22px; \n",
    "              font-family: Arial, sans-serif; \n",
    "              line-height: 1.6em; \n",
    "              text-align: center; \n",
    "              color: #84172D; \n",
    "              margin: 0;\">\n",
    "        🙏 <b>Thank you for reading this notebook!</b>\n",
    "    </p>\n",
    "</div>\n"
   ]
  }
 ],
 "metadata": {
  "kaggle": {
   "accelerator": "nvidiaTeslaT4",
   "dataSources": [
    {
     "databundleVersionId": 861823,
     "sourceId": 3004,
     "sourceType": "competition"
    }
   ],
   "dockerImageVersionId": 30762,
   "isGpuEnabled": true,
   "isInternetEnabled": true,
   "language": "python",
   "sourceType": "notebook"
  },
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.14"
  },
  "papermill": {
   "default_parameters": {},
   "duration": 1327.521856,
   "end_time": "2024-09-23T20:52:58.312881",
   "environment_variables": {},
   "exception": null,
   "input_path": "__notebook__.ipynb",
   "output_path": "__notebook__.ipynb",
   "parameters": {},
   "start_time": "2024-09-23T20:30:50.791025",
   "version": "2.6.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}