tensorflow_Implementation.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "######  Necessary library"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "import numpy as np\n",
    "from pylab import rcParams\n",
    "from numpy import linalg as LA\n",
    "import matplotlib.pyplot as plt\n",
    "from keras.datasets import fashion_mnist\n",
    "from keras.utils import to_categorical\n",
    "# Improve progress bar display\n",
    "from tqdm import tqdm\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "tf.enable_eager_execution()\n",
    "tf.set_random_seed(1867)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Batch_Size for training\n",
    "batch_size = 64\n",
    "#number of classes in fashion mnist datasets  \n",
    "num_classes = 10\n",
    "# no of epoch\n",
    "epochs = 10"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### Load Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Loading data using load_data() API\n",
    "(x_train, y_train), (x_test, y_test) = tf.keras.datasets.fashion_mnist.load_data()\n",
    "\n",
    "#tf.data, the recommended API for building input pipelines in TensorFlow.\n",
    "dataset_train = tf.data.Dataset.from_tensor_slices((\n",
    "    tf.cast(x_train/255, tf.float32),\n",
    "    tf.cast(y_train, tf.int64)\n",
    ")).shuffle(1000).batch(64)\n",
    "\n",
    "dataset_test = tf.data.Dataset.from_tensor_slices((\n",
    "    tf.cast(x_test/255, tf.float32),\n",
    "    tf.cast(y_test, tf.int64)\n",
    ")).batch(64)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [],
   "source": [
    "'''\n",
    "    Implements the forward propagation for the model:\n",
    "    LINEAR->RELU->LINEAR->RELU->LINEAR->RELU->LINEAR->RELU->LINEAR->SOFTMAX\n",
    "'''\n",
    "input_shape = x_train[1].shape\n",
    "X_input = tf.keras.Input(input_shape)\n",
    "\n",
    "X = tf.keras.layers.Dense(1000, activation=tf.nn.relu, use_bias=False)(X_input)\n",
    "X = tf.keras.layers.Dense(1000, activation=tf.nn.relu, use_bias=False)(X)\n",
    "X = tf.keras.layers.Dense(500, activation=tf.nn.relu, use_bias=False)(X)\n",
    "X = tf.keras.layers.Dense(200, activation=tf.nn.relu, use_bias=False)(X)\n",
    "X = tf.keras.layers.Dense(num_classes, activation=tf.nn.relu,use_bias=False)(X)\n",
    "\n",
    "fmnist_model=tf.keras.Model(inputs=X_input,name='fmnist_model',outputs=X)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### Training  the model  using eager execution"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "70b251a7fcd74366b31cbab4aaad4d01",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=938), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "146c0b9497d54af68bdebe7e79be6b95",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=938), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b7cbee3eaf4e448395e4ef86197eba2c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=938), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "cc2a6bd8360944fe867ae1b173ddfb3f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=938), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6f1e2f4fea664cc8a1305e308d543106",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=938), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "5aa2f80230d843ae8981ed7dae0d820f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=938), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "60c249dd0b4843c78c81b7fd1f5127c6",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=938), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "5128240436a34a648bc3bac282d5531f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=938), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6332e488250442d79b0152229a1c90f0",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=938), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "461a7b98ab6843da8d7221306188fac8",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=938), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "# Backpropagation: Define the tensorflow optimizer. Use an AdamOptimizer\n",
    "optimizer = tf.train.AdamOptimizer(learning_rate=0.0001)\n",
    "\n",
    "\"\"\"\n",
    "global_step refers to the number of batches seen by the graph. Every time a batch is provided,\n",
    "the weights are updated in the direction that minimizes the loss.\n",
    "global_step just keeps track of the number of batches seen so far.\n",
    "\"\"\"\n",
    "global_step = tf.train.get_or_create_global_step()\n",
    "\n",
    "# For plotting the loss of the model \n",
    "training_losses = []\n",
    "# For plotting the accuracy of the model \n",
    "training_accuracies = []\n",
    "\n",
    "for epoch in range(epochs):\n",
    "    #Loss function \n",
    "    epoch_loss_avg = tf.contrib.eager.metrics.Mean()\n",
    "    #Metric for evalaution of the model \n",
    "    epoch_accuracy = tf.contrib.eager.metrics.Accuracy()\n",
    "    for x, y in tqdm(dataset_train, total=round(len(x_train)/batch_size)):\n",
    "        # For Calculating the graidents for the backpropagation\n",
    "        with tf.GradientTape() as tape:\n",
    "            outputs = model(x)\n",
    "            loss = tf.losses.softmax_cross_entropy(tf.one_hot(y, num_classes), outputs)\n",
    "        grads = tape.gradient(loss, model.trainable_weights)\n",
    "        optimizer.apply_gradients(zip(grads, model.trainable_weights), global_step)\n",
    "        epoch_loss_avg(loss)\n",
    "        epoch_accuracy(tf.argmax(outputs, axis=1, output_type=tf.int64), y)\n",
    "    \n",
    "    training_losses.append(epoch_loss_avg.result())\n",
    "    training_accuracies.append(epoch_accuracy.result())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0,0.5,'Train Accuracy')"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1440x360 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot training curves\n",
    "plt.figure(1, figsize=(20,5))\n",
    "plt.subplot(121)\n",
    "plt.plot(training_losses)\n",
    "plt.title('Train_Loss vs Epochs')\n",
    "plt.xlabel('Epoch')\n",
    "plt.ylabel('Train Loss')\n",
    "\n",
    "plt.subplot(122)\n",
    "plt.plot(training_accuracies)\n",
    "plt.title(' Train_Accuracy vs Epochs')\n",
    "plt.xlabel('Epoch')\n",
    "plt.ylabel('Train Accuracy')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [],
   "source": [
    "def test_evaluate(model, dataset):\n",
    "    \"\"\"\n",
    "    Method for evaluation test data\n",
    "    \n",
    "    Arguments:\n",
    "    model -- tensorflow model\n",
    "    dataset -- Data to evaluate \n",
    "    \n",
    "    Returns:\n",
    "    epoch_loss_avg --Test loss\n",
    "    epoch_accuracy -- Test accuracy\n",
    "    \"\"\"\n",
    "    #Loss \n",
    "    epoch_loss_avg = tf.contrib.eager.metrics.Mean()\n",
    "    #Metric for evalaution of the model \n",
    "    epoch_accuracy = tf.contrib.eager.metrics.Accuracy()\n",
    "    for x, y in dataset:\n",
    "        outputs = model(x)\n",
    "        loss = tf.losses.softmax_cross_entropy(tf.one_hot(y, 10), outputs)\n",
    "        epoch_loss_avg(loss)\n",
    "        epoch_accuracy(tf.argmax(outputs, axis=1, output_type=tf.int64), y)\n",
    "        \n",
    "    return epoch_loss_avg.result().numpy(), epoch_accuracy.result().numpy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x432 with 5 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot histogram of each layer's weights (for weight pruning)\n",
    "plt.figure(1, figsize=(15,6))\n",
    "plt.subplots_adjust(hspace=0.5)\n",
    "for i, weights in enumerate(model.trainable_weights):\n",
    "    plt.subplot(3,2,i+1)\n",
    "    plt.title(\"Layer {}\".format(i+1))\n",
    "    plt.hist(weights.numpy().reshape(-1), bins=100)\n",
    "    plt.xlim(-0.5, 0.5);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x432 with 5 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot histogram of each layer's coloumn norms (for unit pruning)\n",
    "plt.figure(2, figsize=(15,6))\n",
    "plt.subplots_adjust(hspace=0.5)\n",
    "for i, weights in enumerate(model.trainable_weights):\n",
    "    column_norms = np.linalg.norm(weights.numpy(), ord=2, axis=0)\n",
    "    plt.subplot(3,2,i+1)\n",
    "    plt.title(\"Layer {}\".format(i+1))\n",
    "    plt.hist(column_norms, bins=100);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Saving the model weights\n",
    "model.save_weights('pre_trained.h5', save_format='h5')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Weight Pruning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [],
   "source": [
    "def weight_prune(weight, k):\n",
    "    '''\n",
    "    Method to rank the individual weights in weight matrix W\n",
    "    according to their magnitude (absolute value), and then set to zero the smallest\n",
    "    k%.\n",
    "    \n",
    "    Arguments:\n",
    "    weight -- weigths to prune\n",
    "    k -- pruning percent\n",
    "    \n",
    "    Returns:\n",
    "    weights  = pruned weigths\n",
    "    '''\n",
    "    weights = []\n",
    "    weight_temp = weight\n",
    "    weight_temp = np.absolute(weight_temp)\n",
    "    weight_temp = np.sort(weight_temp, axis = None)\n",
    "    threshold = weight_temp[int(k*weight_temp.size)] #Finding the threshold weight\n",
    "    weight[(weight < threshold) & (weight > -threshold)] = 0 #Setting the lowest k% weights to 0\n",
    "    weights.append(weight)\n",
    "\n",
    "    return weights"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#Weight Pruning\n",
    "accuracy_weight_pruning=[]\n",
    "original_accuracy=[]\n",
    "sparsity = [0, 25, 50, 60, 70, 80, 90, 95, 97, 99]\n",
    "\n",
    "for k2 in sparsity:\n",
    "    k = k2*0.01\n",
    "    for i in range(1,5):\n",
    "        # get_weights get the  weights of the particular layer\n",
    "        weight = model.layers[i].get_weights()[0]\n",
    "        # set_ weights set assign prune to weights to particular layer\n",
    "        model.layers[i].set_weights(weight_prune(weight, k))\n",
    "    score, acc = test_evaluate(model, dataset_test)\n",
    "    accuracy_weight_pruning.append(acc*100)\n",
    "    model.load_weights('pre_trained.h5') #Reloading weights\n",
    "    score, acc = test_evaluate(model, dataset_test)\n",
    "    original_accuracy.append(acc*100)\n",
    "    \n",
    "\n",
    "    \n",
    "rcParams['figure.figsize'] = 12, 8\n",
    "plt.plot(sparsity,original_accuracy,'r', sparsity,accuracy_weight_pruning,'b')\n",
    "plt.grid()\n",
    "plt.legend([['Original Accuracy(without pruning)'],['Accuracy after weight pruning']],\n",
    "           loc='lower left', fontsize='xx-large')\n",
    "plt.xlabel('Percentage sparsity',fontsize='xx-large')\n",
    "plt.ylabel('Percentage Accuracy',fontsize='xx-large')\n",
    "plt.xticks(sparsity)\n",
    "plt.yticks(np.arange(0,100,5))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### Unit/Neuron Pruning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [],
   "source": [
    "def unit_prune(weight, k):\n",
    "    '''\n",
    "    Method to rank the columns of a weight matrix according\n",
    "    to their L2-norm and delete the smallest k%.\n",
    "    \n",
    "    Arguments:\n",
    "    weight -- weigths to prune\n",
    "    k -- pruning percent \n",
    "    \n",
    "    Returns:\n",
    "    weights  = pruned weigths\n",
    "    ''' \n",
    "    weights = []\n",
    "    weight_temp = np.linalg.norm(weight, axis=0) #Finding norm of each column\n",
    "    sorted = np.sort(weight_temp)\n",
    "    threshold = sorted[int(k*sorted.size)] #Finding threshold norm\n",
    "    j=0\n",
    "    for i in weight_temp:\n",
    "        if(i<threshold):\n",
    "            weight[:, j] = 0 #set jth column in weight matrix to 0 because it's norm is less than that of threshold\n",
    "        j=j+1\n",
    "    weights.append(weight)\n",
    "    return weights"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#Unit Pruning\n",
    "accuracy_unit_pruning=[]\n",
    "original_accuracy=[]\n",
    "\n",
    "sparsity = [0, 25, 50, 60, 70, 80, 90, 95, 97, 99]\n",
    "\n",
    "for k2 in sparsity:\n",
    "    k = k2*0.01\n",
    "    for i in range(1,5):\n",
    "        # get_weights get the  weights of the particular layer\n",
    "        weight = model.layers[i].get_weights()[0]\n",
    "        # set_ weights set assign prune to weights to particular layer\n",
    "        model.layers[i].set_weights(unit_prune(weight, k))\n",
    "    score, acc = test_evaluate(model, dataset_test)\n",
    "    accuracy_unit_pruning.append(acc*100)\n",
    "    model.load_weights('pre_trained.h5') #Reloading weights\n",
    "    score, acc = test_evaluate(model, dataset_test)\n",
    "    original_accuracy.append(acc*100)\n",
    "\n",
    "    \n",
    "rcParams['figure.figsize'] = 12, 8\n",
    "plt.plot(sparsity,original_accuracy,'r', sparsity,accuracy_unit_pruning,'g')\n",
    "plt.grid()\n",
    "plt.legend([['Original accuracy without pruning'],['Accuracy after Neuron/Unit pruning']],\n",
    "           loc='lower left', fontsize='xx-large')\n",
    "plt.xlabel('Percentage sparsity',fontsize='xx-large')\n",
    "plt.ylabel('Percentage Accuracy',fontsize='xx-large')\n",
    "plt.xticks(sparsity)\n",
    "plt.yticks(np.arange(0,100,5))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Conclusions\n",
    "\n",
    "* we see that the weights of the model (barring those in the final layer) are normally distributed, with most of the weights being centered around 0. Thus when weight pruning, the vast majority of weights are of low magnitude and can be removed with minimal impact to performance. In contrast,histogram of each layer's coloumn norms shows that the column norms of the weights are far more spread out, and not centered around 0 (by definition of the L2 norm). If we view the individual weights in a column as being independent draws from the normal distribution for that layer (maybe not a valid assumption), then the L2 column norms would follow a chi-distribution (which seems in-line with the histograms actually). When pruning columns, it's unlikely that a column is filled with many high-magnitude weights; it's more common for there to be a mixture of low and high magnitude weights.\n",
    "\n",
    "\n",
    "* As the sparsity increases from 0 to 50%, there is a very small decrease in accuracy. This means that at least half of the neurons in a neural network are redundant and do not contribute to accuracy at all (the number of redundant neurons is higher in case of neuron pruning as compared to weight pruning).\n",
    "\n",
    "\n",
    "* In the Weight pruning method, even after pruning 70% of weights in each layer, the accuracy remains at 90%. We can say only 30%(approx. 715,800 out of 2,386,000 parameters) are contributing to the final output. This sparsity can be used to decrease network size and to increase performance.\n",
    "\n",
    "\n",
    "* Neuron pruning shows a significantly faster drop in accuracy as compared to weight pruning. The reason for this is that for a sparsity of k%, k% of the weights become zero in case of weight pruning, whereas in neuron pruning k% of the neurons are deleted in each layer which leads to the deletion of lot more connections as compared to weight pruning with same sparsity.\n",
    "\n",
    "\n",
    "* The presence of an excess of redundant neurons provides scope for decreasing the size of neural network while maintaining the same accuracy."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "TENSORFLOW_GPU",
   "language": "python",
   "name": "tensorflow-gpu"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}