Basic_AutoEncoder.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Basic_AutoEncoder.ipynb",
      "version": "0.3.2",
      "provenance": [],
      "collapsed_sections": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/sayakpaul/TF-2.0-Hacks/blob/master/Basic_AutoEncoder.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xJFE4zpVTsb-",
        "colab_type": "text"
      },
      "source": [
        "The motivation behind this notebook comes from [Sebastian Rachka](https://sebastianraschka.com/) who constantly inspires me through his work in the field of deep learning. Quite a few days, he open-sourced his repo [deeplearning-models](https://github.com/rasbt/deeplearning-models) which contains implementations of a wide variety of deep learning models. I started with [this notebook](https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/autoencoder/ae-basic.ipynb) which shows a very simplistic and minimal implementation of a fully-connected _autoencoder_."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "GAVJ51jU-IDn",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "!pip install tensorflow-gpu==2.0.0-beta0"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "iVlye4JL-WF5",
        "colab_type": "code",
        "outputId": "3b50c843-1325-464e-bb92-d44a5ad4c841",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 36
        }
      },
      "source": [
        "# Imports\n",
        "import tensorflow as tf\n",
        "from tensorflow.keras.datasets import mnist\n",
        "import numpy as np\n",
        "np.random.seed(7)\n",
        "\n",
        "print(tf.__version__)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "2.0.0-beta0\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pkA93pC6-knh",
        "colab_type": "code",
        "outputId": "05da897f-3658-4fb1-9702-4f9b8d9110cd",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 74
        }
      },
      "source": [
        "# Load the MNIST dataset\n",
        "(X_train, y_train), (X_test, y_test) = mnist.load_data()\n",
        "\n",
        "print(X_train.shape, X_test.shape)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/mnist.npz\n",
            "11493376/11490434 [==============================] - 0s 0us/step\n",
            "(60000, 28, 28) (10000, 28, 28)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "W2rHWxLTEPig",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Define the constants\n",
        "NUM_FEATURES = 784\n",
        "UNITS = 32"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zHbkofHY-3Bk",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Custom class for a simple FC autoencoder\n",
        "class AutoEncoder(tf.keras.Model):\n",
        "  \n",
        "  def __init__(self, num_features, units):\n",
        "    super(AutoEncoder, self).__init__()\n",
        "    self.encoder = tf.keras.layers.Dense(units, activation='linear', \n",
        "                                         input_shape=(num_features,),\n",
        "                                         kernel_initializer='glorot_normal',\n",
        "                                         bias_initializer='zeros')\n",
        "    self.decoder = tf.keras.layers.Dense(num_features, activation='linear', input_shape=(units,))\n",
        "    self.leaky_relu = tf.keras.layers.LeakyReLU(0.5)\n",
        "  \n",
        "  def call(self, x):\n",
        "    encoded = self.encoder(x)\n",
        "    encoded = self.leaky_relu(encoded)\n",
        "    decoded = tf.sigmoid(self.decoder(encoded))\n",
        "    return decoded\n",
        "\n",
        "# Instantiate the autoencoder\n",
        "auto_encoder = AutoEncoder(NUM_FEATURES, UNITS)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-6rsrMCcEsY8",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Define loss function and optimizer\n",
        "loss_func = tf.keras.losses.BinaryCrossentropy()\n",
        "optimizer = tf.keras.optimizers.Adam(learning_rate=0.001)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7Ajqnm69FR56",
        "colab_type": "code",
        "outputId": "0dedf00f-70ce-485f-c322-136da53d7197",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 36
        }
      },
      "source": [
        "# Flatten the training images\n",
        "X_train_copy = X_train.copy()\n",
        "\n",
        "X_train_copy = X_train_copy.reshape(60000, 28*28).astype(np.float32)\n",
        "\n",
        "print(X_train_copy.shape)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(60000, 784)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fhXJ-_9aGUTx",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Batches of 64\n",
        "train_ds = tf.data.Dataset.from_tensor_slices((X_train_copy, y_train)).batch(64)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "awOBPf9pHG8a",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Average out the loss after each epoch\n",
        "train_loss = tf.keras.metrics.Mean(name='train_loss')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Si85usXjG8Ve",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Train the model\n",
        "@tf.function\n",
        "def model_train(features):\n",
        "    # Define the GradientTape context\n",
        "    with tf.GradientTape() as tape:\n",
        "        # Get the probabilities\n",
        "        decoded = auto_encoder(features)\n",
        "        # Calculate the loss\n",
        "        loss = loss_func(decoded, features)\n",
        "    # Get the gradients\n",
        "    gradients = tape.gradient(loss, auto_encoder.trainable_variables)\n",
        "    # Update the weights\n",
        "    optimizer.apply_gradients(zip(gradients, auto_encoder.trainable_variables))\n",
        "\n",
        "    train_loss(loss)\n",
        "    return decoded"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "390JgOXiHe2h",
        "colab_type": "code",
        "outputId": "2ca000e9-ddb8-4e48-d0db-f16760fe1953",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 397
        }
      },
      "source": [
        "# Begin training\n",
        "decode_list = []\n",
        "for epoch in range(20):\n",
        "  for features, _ in train_ds:\n",
        "    features = features\n",
        "    decoded = model_train(features)\n",
        "  \n",
        "  template = 'Epoch {}, loss: {}'\n",
        "  print (template.format(epoch+1,\n",
        "                         train_loss.result()))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1, loss: 1.649407982826233\n",
            "Epoch 2, loss: 1.3055578470230103\n",
            "Epoch 3, loss: 1.1296701431274414\n",
            "Epoch 4, loss: 1.0245317220687866\n",
            "Epoch 5, loss: 0.9548731446266174\n",
            "Epoch 6, loss: 0.9051530957221985\n",
            "Epoch 7, loss: 0.866610586643219\n",
            "Epoch 8, loss: 0.8354974389076233\n",
            "Epoch 9, loss: 0.8103290796279907\n",
            "Epoch 10, loss: 0.7897171378135681\n",
            "Epoch 11, loss: 0.7726050019264221\n",
            "Epoch 12, loss: 0.7579894065856934\n",
            "Epoch 13, loss: 0.7453299760818481\n",
            "Epoch 14, loss: 0.7341734170913696\n",
            "Epoch 15, loss: 0.7243269085884094\n",
            "Epoch 16, loss: 0.715560793876648\n",
            "Epoch 17, loss: 0.7077820897102356\n",
            "Epoch 18, loss: 0.700776994228363\n",
            "Epoch 19, loss: 0.6943405866622925\n",
            "Epoch 20, loss: 0.6884656548500061\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BYWy0mc7H29S",
        "colab_type": "code",
        "outputId": "771160e3-8769-4de5-9365-b7fd4abfe413",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 187
        }
      },
      "source": [
        "%matplotlib inline\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "##########################\n",
        "### VISUALIZATION\n",
        "##########################\n",
        "\n",
        "n_images = 15\n",
        "image_width = 28\n",
        "\n",
        "fig, axes = plt.subplots(nrows=2, ncols=n_images, \n",
        "                         sharex=True, sharey=True, figsize=(20, 2.5))\n",
        "orig_images = features[:n_images].numpy()\n",
        "decoded_images = decoded[:n_images].numpy()\n",
        "\n",
        "for i in range(n_images):\n",
        "    for ax, img in zip(axes, [orig_images, decoded_images]):\n",
        "        curr_img = img[i]\n",
        "        ax[i].imshow(curr_img.reshape((image_width, image_width)), cmap='binary')"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAABIAAAACqCAYAAADV0DQZAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 1440x180 with 30 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    }
  ]
}