Generative Adversarial Network.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/mayank/anaconda2/lib/python2.7/site-packages/matplotlib/__init__.py:872: UserWarning: axes.color_cycle is deprecated and replaced with axes.prop_cycle; please use the latter.\n",
      "  warnings.warn(self.msg_depr % (key, alt_key))\n",
      "Using gpu device 0: GeForce GTX 960M (CNMeM is disabled)\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "import theano.tensor as T\n",
    "import lasagne"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import theano"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from scipy.stats import norm\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "mu, sigma = -1,1\n",
    "xs = np.linspace(-5,5,1000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x7fc5e8787490>]"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAAEDCAYAAAA7jc+ZAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fc5e8839810>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(xs, norm.pdf(xs, loc=mu, scale=sigma))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "train_iters = 500\n",
    "M = 200"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "input_var_d = T.col('input_d')\n",
    "target_var_d = T.col('output_d')\n",
    "input_var_g = T.col('input_g')\n",
    "target_var_g = T.col('output_g')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 123,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "#### Pre-Training\n",
    "\n",
    "##Discriminator\n",
    "\n",
    "l_in = lasagne.layers.InputLayer(shape=(M, 1), input_var=input_var_d)\n",
    "l_fw_1 = lasagne.layers.DenseLayer(l_in, num_units = 6, nonlinearity=lasagne.nonlinearities.tanh,\n",
    "                                  W = lasagne.init.GlorotUniform(),\n",
    "                                  b = lasagne.init.Constant(0.0))\n",
    "l_fw_2 = lasagne.layers.DenseLayer(l_fw_1, num_units = 5, nonlinearity=lasagne.nonlinearities.tanh,\n",
    "                                  W = lasagne.init.GlorotUniform(),\n",
    "                                  b = lasagne.init.Constant(0.0))\n",
    "l_out = lasagne.layers.DenseLayer(l_fw_2, num_units = 1, nonlinearity =lasagne.nonlinearities.tanh,\n",
    "                                 W = lasagne.init.GlorotUniform(),\n",
    "                                 b = lasagne.init.Constant(0.0))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 124,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "prediction = lasagne.layers.get_output(l_out)\n",
    "loss = lasagne.objectives.squared_error(prediction, target_var_d)\n",
    "loss = loss.mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 125,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "params = lasagne.layers.get_all_params(l_out, trainable=True)\n",
    "updates = lasagne.updates.momentum(loss, params, learning_rate = 0.03)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 126,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "train = theano.function([input_var_d, target_var_d], loss, updates=updates, allow_input_downcast=True)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 127,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "output = theano.function([input_var_d], prediction, allow_input_downcast=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 128,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "###Plotting the decision surface\n",
    "def plot_d0():\n",
    "    f, ax = plt.subplots(1)\n",
    "    xs = np.linspace(-5,5,1000)\n",
    "    ax.plot(xs, norm.pdf(xs, loc=mu, scale=sigma), label='p_data')\n",
    "    \n",
    "    ###Decision Boundry\n",
    "    r = 1000 #Resolution\n",
    "    xs = np.linspace(-5,5,r)\n",
    "    ds = np.zeros((r,1)) #Decision Surface\n",
    "    \n",
    "    ####We process multiple points in parallel in a minibatch\n",
    "    for i in range(r/M):\n",
    "        x = np.reshape(xs[M*i:M*(i+1)], (M,1))\n",
    "        ds[M*i:M*(i+1)] = output(x)\n",
    "    \n",
    "    ax.plot(xs, ds, label='decision boundry')\n",
    "    ax.set_ylim(0,1.1)\n",
    "    plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 129,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.text.Text at 0x7fc5dfcd6150>"
      ]
     },
     "execution_count": 129,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXUAAAEMCAYAAAA70CbBAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nBz6y4nstCDAGAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fc5d8bbdf10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_d0()\n",
    "plt.title('Initial Decision Boundry')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 130,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0\n",
      "100\n",
      "200\n",
      "300\n",
      "400\n",
      "500\n",
      "600\n",
      "700\n",
      "800\n",
      "900\n"
     ]
    }
   ],
   "source": [
    "## Pretraining starts\n",
    "\n",
    "lh = np.zeros(1000)\n",
    "for i in range(1000):\n",
    "    d = (np.random.random(M))\n",
    "    d = np.reshape(d, (M,1))\n",
    "    labels =norm.pdf(d, loc=mu, scale=sigma)\n",
    "    labels = np.reshape(labels, (M,1))\n",
    "    lh[i] = train(d, labels)\n",
    "    if i%100 == 0:\n",
    "        print i"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 131,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x7fc5da930450>]"
      ]
     },
     "execution_count": 131,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX0AAAEDCAYAAADZUdTgAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fc5db3c7850>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(lh)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 132,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXUAAAD+CAYAAADF/ZVnAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ncPXpC+nclB8AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fc5da4dfe50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_d0()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 193,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "##### The actual generative adversarial network #######\n",
    "\n",
    "## Generator\n",
    "g_in = lasagne.layers.InputLayer(shape=(M,1), input_var=input_var_g)\n",
    "g_fw_1 = lasagne.layers.DenseLayer(g_in, num_units=6, nonlinearity=lasagne.nonlinearities.tanh,\n",
    "                                  W = lasagne.init.GlorotUniform(),\n",
    "                                  b = lasagne.init.Constant(0.0))\n",
    "g_fw_2 = lasagne.layers.DenseLayer(g_fw_1, num_units=5, nonlinearity=lasagne.nonlinearities.tanh,\n",
    "                                  W = lasagne.init.GlorotUniform(),\n",
    "                                  b = lasagne.init.Constant(0.0))\n",
    "g_out = lasagne.layers.DenseLayer(g_fw_2, num_units=1, nonlinearity=lasagne.nonlinearities.tanh,\n",
    "                                  W = lasagne.init.GlorotUniform(),\n",
    "                                  b = lasagne.init.Constant(0.0))\n",
    "\n",
    "#Generated_sample\n",
    "prediction_g = lasagne.layers.get_output(g_out)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 194,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "## Discriminator - D(G(x))\n",
    "dg_fw_1 = lasagne.layers.DenseLayer(g_out, num_units = 6, nonlinearity=lasagne.nonlinearities.tanh,\n",
    "                                  W = lasagne.init.GlorotUniform(),\n",
    "                                  b = lasagne.init.Constant(0.0))\n",
    "dg_fw_2 = lasagne.layers.DenseLayer(dg_fw_1, num_units = 5, nonlinearity=lasagne.nonlinearities.tanh,\n",
    "                                  W = lasagne.init.GlorotUniform(),\n",
    "                                  b = lasagne.init.Constant(0.0))\n",
    "dg_out = lasagne.layers.DenseLayer(dg_fw_2, num_units = 1, nonlinearity =lasagne.nonlinearities.tanh,\n",
    "                                 W = lasagne.init.GlorotUniform(),\n",
    "                                 b = lasagne.init.Constant(0.0))\n",
    "\n",
    "#Prediction\n",
    "prediction_dg = lasagne.layers.get_output(dg_out)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 195,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "## Discriminator - D((x))\n",
    "#Shares weights with D(G(x))\n",
    "d_in = lasagne.layers.InputLayer(shape=(M, 1), input_var=input_var_d)\n",
    "d_fw_1 = lasagne.layers.DenseLayer(d_in, num_units = 6, nonlinearity=lasagne.nonlinearities.tanh,\n",
    "                                  W = dg_fw_1.W,\n",
    "                                  b = dg_fw_1.b)\n",
    "d_fw_2 = lasagne.layers.DenseLayer(d_fw_1, num_units = 5, nonlinearity=lasagne.nonlinearities.tanh,\n",
    "                                  W = dg_fw_2.W,\n",
    "                                  b = dg_fw_2.b)\n",
    "d_out = lasagne.layers.DenseLayer(d_fw_2, num_units = 1, nonlinearity =lasagne.nonlinearities.tanh,\n",
    "                                 W = dg_out.W,\n",
    "                                 b = dg_out.b)\n",
    "\n",
    "#Prediction\n",
    "prediction_d = lasagne.layers.get_output(d_out)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 196,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "params_d_g_values = lasagne.layers.get_all_param_values(dg_out)\n",
    "\n",
    "params_d_values = lasagne.layers.get_all_param_values(d_out)\n",
    "\n",
    "params_g_values = lasagne.layers.get_all_param_values(g_out)\n",
    "\n",
    "params_pretrained_d_values = lasagne.layers.get_all_param_values(l_out)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 197,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "## Using pretrained weights to imporve D\n",
    "\n",
    "lasagne.layers.set_all_param_values(d_out, params_pretrained_d_values)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 198,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "## Getting the parameters\n",
    "params_d_g = lasagne.layers.get_all_params(dg_out)\n",
    "\n",
    "params_g = lasagne.layers.get_all_params(g_out)\n",
    "\n",
    "params_d = lasagne.layers.get_all_params(d_out)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 199,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "## Objectives\n",
    "obj_d = T.mean(T.log(prediction_d) + T.log(1-prediction_dg))\n",
    "obj_g = T.mean(T.log(prediction_dg))\n",
    "\n",
    "## Updates\n",
    "updates_d = lasagne.updates.momentum(1-obj_d, params_d, learning_rate = 0.01)\n",
    "updates_g = lasagne.updates.momentum(1-obj_g, params_d_g, learning_rate = 0.01)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 200,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "## Train functions ##\n",
    "train_d = theano.function([input_var_g, input_var_d], obj_d, updates=updates_d, allow_input_downcast=True)\n",
    "\n",
    "train_g = theano.function([input_var_g], obj_g, updates=updates_g, allow_input_downcast=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 201,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "## Output functions##\n",
    "out_d = theano.function([input_var_d], prediction_d, allow_input_downcast=True)\n",
    "out_dg = theano.function([input_var_g], prediction_dg, allow_input_downcast=True)\n",
    "out_g = theano.function([input_var_g], prediction_g, allow_input_downcast=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 202,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "###### MULTIPLY GEN OUT BY 5######\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "def plot_fig():\n",
    "    \n",
    "    f, ax = plt.subplots(1)\n",
    "    #p_data\n",
    "    xs = np.linspace(-5,5,1000)\n",
    "    ax.plot(xs, norm.pdf(xs, loc=mu, scale=sigma), label='p_data')\n",
    "    \n",
    "    #decision_boundry\n",
    "    r=5000\n",
    "    xs = np.linspace(-5,5,r)\n",
    "    ds=np.zeros((r,1))\n",
    "    #process muliple points in parallel\n",
    "    for i in range(r/M):\n",
    "        x=np.reshape(xs[M*i:M*(i+1)], (M,1))\n",
    "        ds[M*i:M*(i+1)]=out_d(x)\n",
    "        \n",
    "    ax.plot(xs, ds, label='decision_boundry')\n",
    "        \n",
    "   \n",
    "    \n",
    "    \n",
    "    \n",
    "    # distribution of inverse mapped points\n",
    "    zs = np.linspace(-5,5,r)\n",
    "    gs = np.zeros((r,1))\n",
    "    for i in range(r/M):\n",
    "        z=np.reshape(zs[M*i:M*(i+1)],(M,1))\n",
    "        gs[M*i:M*(i+1)]=out_g(z)\n",
    "    histc, edges = np.histogram(gs, bins=10)\n",
    "    ax.plot(np.linspace(-5,5,10), histc/float(r), label='p_g')\n",
    "    \n",
    "    ax.set_ylim(-2,1.1)\n",
    "    plt.legend()\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 203,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAAD+CAYAAAA09s7qAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nhRDC4STRCyGEw/1/t2YHU8AkaiEAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fc5df1b7250>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_fig()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 204,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.0\n",
      "0.1\n",
      "0.2\n",
      "0.3\n",
      "0.4\n",
      "0.5\n",
      "0.6\n",
      "0.7\n",
      "0.8\n",
      "0.9\n"
     ]
    }
   ],
   "source": [
    "#Training algo\n",
    "k=1\n",
    "histd, histg = np.zeros(train_iters), np.zeros(train_iters)\n",
    "\n",
    "for i in range(train_iters):\n",
    "    for j in range(k):\n",
    "        x = np.random.normal(mu, sigma, M)\n",
    "        x.sort()\n",
    "        z = np.linspace(-5.0,5.0,M)+np.random.random(M)*0.01\n",
    "        histd[i] = train_d(np.reshape(z,(M,1)), np.reshape(x,(M,1)))\n",
    "        \n",
    "    z = np.linspace(-5.0,5.0,M)+np.random.random(M)*0.01\n",
    "    histg[i] = train_g(np.reshape(z,(M,1)))\n",
    "    if i%(train_iters//10) == 0:\n",
    "        print(float(i)/float(train_iters))\n",
    "        "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 205,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x7fc5d77a6c10>"
      ]
     },
     "execution_count": 205,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXgAAAEDCAYAAAA/eB+kAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nSEgp4EVEQkoBLyISUv8fQ0lgQ4DRXSMAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fc5e0928c90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(histd, label='obj_d')\n",
    "plt.plot(1-histg, label='obj_g')\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 206,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAAD+CAYAAAA09s7qAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fc5d77e6290>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_fig()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.11"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}