VAE-Lab5-CA-VAE.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## VAE-Experiment-5:\n",
    "\n",
    "\n",
    "#### Conditional Autoencoder"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In the previous notebook, we trained an Adversarial Autoencoder (AAE) on unsupervised data. We can also train an AAE with additional label information as input. This allows us to train a conditional generator that can generate samples corresponding to specific input conditionals.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### Experiment\n",
    "Let's now learn an Adversarial Autoencoder that can generate specific digits similar to the ones in the MNIST dataset in a controlled manner.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "##########################\n",
    "# Import necessary modules\n",
    "##########################\n",
    "\n",
    "import pickle\n",
    "import numpy as np\n",
    "import matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "from matplotlib import gridspec\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')\n",
    "%matplotlib inline\n",
    "\n",
    "import torch\n",
    "from torch.autograd import Variable\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "import torch.optim as optim"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "################\n",
    "# Set parameters\n",
    "################\n",
    "\n",
    "cuda = torch.cuda.is_available()\n",
    "kwargs = {'num_workers': 1, 'pin_memory': True} if cuda else {}\n",
    "\n",
    "seed = 10\n",
    "\n",
    "n_classes = 10\n",
    "z_dim = 2\n",
    "X_dim = 784\n",
    "y_dim = 10\n",
    "train_batch_size = 100\n",
    "valid_batch_size = train_batch_size\n",
    "N = 1000\n",
    "epochs = 5\n",
    "\n",
    "params = {}\n",
    "params['cuda'] = cuda\n",
    "params['n_classes'] = n_classes\n",
    "params['z_dim'] = z_dim\n",
    "params['X_dim'] = X_dim\n",
    "params['y_dim'] = y_dim\n",
    "params['train_batch_size'] = train_batch_size\n",
    "params['valid_batch_size'] = valid_batch_size\n",
    "params['N'] = N\n",
    "params['epochs'] = epochs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "###################################\n",
    "# Load data and create Data loaders\n",
    "###################################\n",
    "\n",
    "def load_data(data_path='./data/VAE/processed_data/processed_data/'):\n",
    "    print('loading data!')\n",
    "    trainset_labeled = pickle.load(open(data_path + \"train_labeled.p\", \"rb\"))\n",
    "    trainset_unlabeled = pickle.load(open(data_path + \"train_unlabeled.p\", \"rb\"))\n",
    "    # Set -1 as labels for unlabeled data\n",
    "    trainset_unlabeled.train_labels = torch.from_numpy(np.array([-1] * 47000))\n",
    "    validset = pickle.load(open(data_path + \"validation.p\", \"rb\"))\n",
    "\n",
    "    train_labeled_loader = torch.utils.data.DataLoader(trainset_labeled,\n",
    "                                                       batch_size=train_batch_size,\n",
    "                                                       shuffle=True, **kwargs)\n",
    "\n",
    "    train_unlabeled_loader = torch.utils.data.DataLoader(trainset_unlabeled,\n",
    "                                                         batch_size=train_batch_size,\n",
    "                                                         shuffle=True, **kwargs)\n",
    "\n",
    "    valid_loader = torch.utils.data.DataLoader(validset, batch_size=valid_batch_size, shuffle=True)\n",
    "\n",
    "    return train_labeled_loader, train_unlabeled_loader, valid_loader"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#################\n",
    "# Define Networks\n",
    "#################\n",
    "\n",
    "# Encoder\n",
    "class Q_net(nn.Module):\n",
    "    def __init__(self):\n",
    "        super(Q_net, self).__init__()\n",
    "        self.lin1 = nn.Linear(X_dim, N)\n",
    "        self.lin2 = nn.Linear(N, N)\n",
    "        # Gaussian code (z)\n",
    "        self.lin3gauss = nn.Linear(N, z_dim)\n",
    "\n",
    "    def forward(self, x):\n",
    "        x = F.dropout(self.lin1(x), p=0.2, training=self.training)\n",
    "        x = F.relu(x)\n",
    "        x = F.dropout(self.lin2(x), p=0.2, training=self.training)\n",
    "        x = F.relu(x)\n",
    "        xgauss = self.lin3gauss(x)\n",
    "\n",
    "        return xgauss\n",
    "\n",
    "\n",
    "# Decoder\n",
    "class P_net(nn.Module):\n",
    "    def __init__(self):\n",
    "        super(P_net, self).__init__()\n",
    "        self.lin1 = nn.Linear(z_dim + n_classes, N)\n",
    "        self.lin2 = nn.Linear(N, N)\n",
    "        self.lin3 = nn.Linear(N, X_dim)\n",
    "\n",
    "    def forward(self, x):\n",
    "        x = self.lin1(x)\n",
    "        x = F.dropout(x, p=0.2, training=self.training)\n",
    "        x = F.relu(x)\n",
    "        x = self.lin2(x)\n",
    "        x = F.dropout(x, p=0.2, training=self.training)\n",
    "        x = self.lin3(x)\n",
    "        return F.sigmoid(x)\n",
    "\n",
    "\n",
    "class D_net_gauss(nn.Module):\n",
    "    def __init__(self):\n",
    "        super(D_net_gauss, self).__init__()\n",
    "        self.lin1 = nn.Linear(z_dim, N)\n",
    "        self.lin2 = nn.Linear(N, N)\n",
    "        self.lin3 = nn.Linear(N, 1)\n",
    "\n",
    "    def forward(self, x):\n",
    "        x = F.dropout(self.lin1(x), p=0.2, training=self.training)\n",
    "        x = F.relu(x)\n",
    "        x = F.dropout(self.lin2(x), p=0.2, training=self.training)\n",
    "        x = F.relu(x)\n",
    "\n",
    "        return F.sigmoid(self.lin3(x))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "###################\n",
    "# Utility functions\n",
    "###################\n",
    "\n",
    "def save_model(model, filename):\n",
    "    print('Best model so far, saving it...')\n",
    "    torch.save(model.state_dict(), filename)\n",
    "\n",
    "\n",
    "def report_loss(epoch, D_loss_gauss, G_loss, recon_loss):\n",
    "    '''\n",
    "    Print loss\n",
    "    '''\n",
    "    print('Epoch-{}; D_loss_gauss: {:.4}; G_loss: {:.4}; recon_loss: {:.4}'.format(epoch,\n",
    "                                                                                   D_loss_gauss.data[0],\n",
    "                                                                                   G_loss.data[0],\n",
    "                                                                                   recon_loss.data[0]))\n",
    "\n",
    "\n",
    "def create_latent(Q, loader):\n",
    "    '''\n",
    "    Creates the latent representation for the samples in loader\n",
    "    return:\n",
    "        z_values: numpy array with the latent representations\n",
    "        labels: the labels corresponding to the latent representations\n",
    "    '''\n",
    "    Q.eval()\n",
    "    labels = []\n",
    "\n",
    "    for batch_idx, (X, target) in enumerate(loader):\n",
    "\n",
    "        X = X * 0.3081 + 0.1307\n",
    "        # X.resize_(loader.batch_size, X_dim)\n",
    "        X, target = Variable(X), Variable(target)\n",
    "        labels.extend(target.data.tolist())\n",
    "        if cuda:\n",
    "            X, target = X.cuda(), target.cuda()\n",
    "        # Reconstruction phase\n",
    "        z_sample = Q(X)\n",
    "        if batch_idx > 0:\n",
    "            z_values = np.concatenate((z_values, np.array(z_sample.data.tolist())))\n",
    "        else:\n",
    "            z_values = np.array(z_sample.data.tolist())\n",
    "    labels = np.array(labels)\n",
    "\n",
    "    return z_values, labels\n",
    "\n",
    "\n",
    "def get_categorical(labels, n_classes=10):\n",
    "    cat = np.array(labels.data.tolist())\n",
    "    cat = np.eye(n_classes)[cat].astype('float32')\n",
    "    cat = torch.from_numpy(cat)\n",
    "    return Variable(cat)\n",
    "\n",
    "\n",
    "def get_X_batch(data_loader, params, size=None):\n",
    "    if size is None:\n",
    "        size = data_loader.batch_size\n",
    "\n",
    "    data_loader.batch_size = size\n",
    "\n",
    "    for X, target in data_loader:\n",
    "        break\n",
    "\n",
    "    train_batch_size = params['train_batch_size']\n",
    "    X_dim = params['X_dim']\n",
    "    cuda = params['cuda']\n",
    "\n",
    "    X = X * 0.3081 + 0.1307\n",
    "\n",
    "    X = X[:size]\n",
    "    target = target[:size]\n",
    "\n",
    "    X.resize_(size, X_dim)\n",
    "    X, target = Variable(X), Variable(target)\n",
    "\n",
    "    if cuda:\n",
    "        X, target = X.cuda(), target.cuda()\n",
    "\n",
    "    return X, target"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "####################\n",
    "# Training procedure\n",
    "####################\n",
    "\n",
    "def train(P, Q, D_gauss, P_decoder, Q_encoder, Q_generator, D_gauss_solver, data_loader):\n",
    "    '''\n",
    "    Train procedure for one epoch.\n",
    "    '''\n",
    "    TINY = 1e-15\n",
    "    # Set the networks in train mode (apply dropout when needed)\n",
    "    Q.train()\n",
    "    P.train()\n",
    "    D_gauss.train()\n",
    "\n",
    "    # Loop through the labeled and unlabeled dataset getting one batch of samples from each\n",
    "    # The batch size has to be a divisor of the size of the dataset or it will return\n",
    "    # invalid samples\n",
    "    for X, target in data_loader:\n",
    "\n",
    "        # Load batch and normalize samples to be between 0 and 1\n",
    "        X = X * 0.3081 + 0.1307\n",
    "        X.resize_(train_batch_size, X_dim)\n",
    "        X, target = Variable(X), Variable(target)\n",
    "        if cuda:\n",
    "            X, target = X.cuda(), target.cuda()\n",
    "\n",
    "        # Init gradients\n",
    "        P.zero_grad()\n",
    "        Q.zero_grad()\n",
    "        D_gauss.zero_grad()\n",
    "\n",
    "\n",
    "        # Reconstruction phase\n",
    "\n",
    "        z_gauss = Q(X)\n",
    "        z_cat = get_categorical(target, n_classes=10)\n",
    "        if cuda:\n",
    "            z_cat = z_cat.cuda()\n",
    "\n",
    "        z_sample = torch.cat((z_cat, z_gauss), 1)\n",
    "\n",
    "        X_sample = P(z_sample)\n",
    "        recon_loss = F.binary_cross_entropy(X_sample + TINY, X.resize(train_batch_size, X_dim) + TINY)\n",
    "\n",
    "        recon_loss.backward()\n",
    "        P_decoder.step()\n",
    "        Q_encoder.step()\n",
    "\n",
    "        P.zero_grad()\n",
    "        Q.zero_grad()\n",
    "        D_gauss.zero_grad()\n",
    "\n",
    "\n",
    "        # Regularization phase\n",
    "\n",
    "        # Discriminator\n",
    "        Q.eval()\n",
    "        z_real_gauss = Variable(torch.randn(train_batch_size, z_dim) * 5.)\n",
    "        if cuda:\n",
    "            z_real_gauss = z_real_gauss.cuda()\n",
    "\n",
    "        z_fake_gauss = Q(X)\n",
    "\n",
    "        D_real_gauss = D_gauss(z_real_gauss)\n",
    "        D_fake_gauss = D_gauss(z_fake_gauss)\n",
    "\n",
    "        D_loss = -torch.mean(torch.log(D_real_gauss + TINY) + torch.log(1 - D_fake_gauss + TINY))\n",
    "\n",
    "        D_loss.backward()\n",
    "        D_gauss_solver.step()\n",
    "\n",
    "        P.zero_grad()\n",
    "        Q.zero_grad()\n",
    "        D_gauss.zero_grad()\n",
    "\n",
    "        # Generator\n",
    "        Q.train()\n",
    "        z_fake_gauss = Q(X)\n",
    "\n",
    "        D_fake_gauss = D_gauss(z_fake_gauss)\n",
    "        G_loss = -torch.mean(torch.log(D_fake_gauss + TINY))\n",
    "\n",
    "        G_loss.backward()\n",
    "        Q_generator.step()\n",
    "\n",
    "        P.zero_grad()\n",
    "        Q.zero_grad()\n",
    "        D_gauss.zero_grad()\n",
    "\n",
    "    return D_loss, G_loss, recon_loss\n",
    "\n",
    "\n",
    "def generate_model(train_labeled_loader, train_unlabeled_loader, valid_loader):\n",
    "    torch.manual_seed(10)\n",
    "\n",
    "    if cuda:\n",
    "        Q = Q_net().cuda()\n",
    "        P = P_net().cuda()\n",
    "        D_gauss = D_net_gauss().cuda()\n",
    "    else:\n",
    "        Q = Q_net()\n",
    "        P = P_net()\n",
    "        D_gauss = D_net_gauss()\n",
    "\n",
    "    # Set learning rates\n",
    "    gen_lr = 0.0001\n",
    "    reg_lr = 0.00005\n",
    "\n",
    "    # Set optimizators\n",
    "    P_decoder = optim.Adam(P.parameters(), lr=gen_lr)\n",
    "    Q_encoder = optim.Adam(Q.parameters(), lr=gen_lr)\n",
    "\n",
    "    Q_generator = optim.Adam(Q.parameters(), lr=reg_lr)\n",
    "    D_gauss_solver = optim.Adam(D_gauss.parameters(), lr=reg_lr)\n",
    "\n",
    "    for epoch in range(epochs):\n",
    "        D_loss_gauss, G_loss, recon_loss = train(P, Q, D_gauss, P_decoder, Q_encoder,\n",
    "                                                 Q_generator,\n",
    "                                                 D_gauss_solver,\n",
    "                                                 valid_loader)\n",
    "        if epoch % 1 == 0:\n",
    "            report_loss(epoch, D_loss_gauss, G_loss, recon_loss)\n",
    "\n",
    "    return Q, P"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "loading data!\n",
      "3000\n",
      "750\n",
      "Epoch-0; D_loss_gauss: 0.7161; G_loss: 1.632; recon_loss: 0.2603\n",
      "Epoch-1; D_loss_gauss: 0.7994; G_loss: 1.597; recon_loss: 0.2324\n",
      "Epoch-2; D_loss_gauss: 0.9326; G_loss: 1.321; recon_loss: 0.2223\n",
      "Epoch-3; D_loss_gauss: 1.016; G_loss: 1.133; recon_loss: 0.2061\n",
      "Epoch-4; D_loss_gauss: 1.059; G_loss: 1.105; recon_loss: 0.2149\n"
     ]
    }
   ],
   "source": [
    "##########################\n",
    "# Train a generative model\n",
    "##########################\n",
    "\n",
    "train_labeled_loader, train_unlabeled_loader, valid_loader = load_data()\n",
    "Q, P = generate_model(train_labeled_loader, train_unlabeled_loader, valid_loader)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "####################\n",
    "# Save trained model\n",
    "####################\n",
    "\n",
    "# Save trained model\n",
    "# torch.save(Q,'./data/VAE/TrainedModels/SAAE_mytraining_Q.pt')\n",
    "# torch.save(P,'./data/VAE/TrainedModels/SAAE_mytraining_P.pt')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "####################\n",
    "# Load trained model\n",
    "####################\n",
    "\n",
    "# Load model trained for 500 epochs\n",
    "Q_pt = torch.load('./data/VAE/TrainedModels/SAAE_preTrained_Q.pt',map_location=lambda storage, location: storage)\n",
    "P_pt = torch.load('./data/VAE/TrainedModels/SAAE_preTrained_P.pt',map_location=lambda storage, location: storage)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAAC7CAYAAAB1qmWGAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nCBZ0IqJAsKATEQWCBZ2IKBAs6EREgfh/Ydk4/VSlopMAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7ff61ed61470>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "##########################\n",
    "# Visualize reconstruction\n",
    "##########################\n",
    "\n",
    "def create_reconstruction(Q, P, data_loader, params):\n",
    "    Q.eval()\n",
    "    P.eval()\n",
    "    X, label = get_X_batch(data_loader, params, size=1)\n",
    "\n",
    "    z_g = Q(X)\n",
    "    z_c = get_categorical(label, n_classes=10)\n",
    "    if cuda:\n",
    "        z_c = z_c.cuda()\n",
    "    z = torch.cat((z_c, z_g), 1)\n",
    "    x = P(z)\n",
    "\n",
    "    img_orig = np.array(X[0].data.tolist()).reshape(28, 28)\n",
    "    img_rec = np.array(x[0].data.tolist()).reshape(28, 28)\n",
    "    plt.subplot(1, 2, 1)\n",
    "    plt.imshow(img_orig)\n",
    "    plt.subplot(1, 2, 2)\n",
    "    plt.imshow(img_rec)\n",
    "\n",
    "\n",
    "data_loader = valid_loader    # Training data:  train_unlabeled_loader  |  Validation data:  valid_loader\n",
    "\n",
    "create_reconstruction(Q_pt, P_pt, data_loader, params)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWQAAADuCAYAAAAOR30qAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nEEQhC4IgJARRyIIgCAnhO3lz6EkPYzrKAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7ff61432a6d8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "######################\n",
    "# Visualize generation\n",
    "######################\n",
    "\n",
    "def grid_plot(Q, P, data_loader, params):\n",
    "    Q.eval()\n",
    "    P.eval()\n",
    "    X,_ = get_X_batch(data_loader, params, size=10)\n",
    "    z_g = Q(X)\n",
    "\n",
    "    n_classes = params['n_classes']\n",
    "    cuda = params['cuda']\n",
    "    z_dim = params['z_dim']\n",
    "\n",
    "    z_cat = np.arange(0, n_classes)\n",
    "    z_cat = np.eye(n_classes)[z_cat].astype('float32')\n",
    "    z_cat = torch.from_numpy(z_cat)\n",
    "    z_cat = Variable(z_cat)\n",
    "    if cuda:\n",
    "        z_cat = z_cat.cuda()\n",
    "\n",
    "    nx, ny = 5, n_classes\n",
    "    plt.subplot()\n",
    "    gs = gridspec.GridSpec(nx, ny, hspace=0.05, wspace=0.05)\n",
    "\n",
    "    for i, g in enumerate(gs):\n",
    "        z_gauss = z_g[i // ny].resize(1, z_dim)\n",
    "        z_gauss0 = z_g[i // ny].resize(1, z_dim)\n",
    "\n",
    "        for _ in range(n_classes - 1):\n",
    "            z_gauss = torch.cat((z_gauss, z_gauss0), 0)\n",
    "\n",
    "        z = torch.cat((z_cat, z_gauss), 1)\n",
    "        x = P(z)\n",
    "\n",
    "        ax = plt.subplot(g)\n",
    "        img = np.array(x[i % ny].data.tolist()).reshape(28, 28)\n",
    "        ax.imshow(img, )\n",
    "        ax.set_xticks([])\n",
    "        ax.set_yticks([])\n",
    "        ax.set_aspect('auto')\n",
    "\n",
    "\n",
    "grid_plot(Q_pt, P_pt, data_loader, params)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Exercise\n",
    "\n",
    "I can feel your excitement :) Check out the research paper on Conditional AutoEncoders: <a href='http://papers.nips.cc/paper/5775-learning-structured-output-representation-using-deep-conditional-generative-models'>Learning structured output representation using deep conditional generative models</a>, NIPS 2015."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "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": "3.6.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}