Visualization.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Visualizing NTM performance on copy task"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "import os\n",
    "import sys\n",
    "import itertools\n",
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.patches as mpatches\n",
    "import matplotlib.gridspec as gridspec\n",
    "from mpl_toolkits.axes_grid1 import make_axes_locatable\n",
    "\n",
    "from ntm import NTM\n",
    "from recurrent_controller import RecurrentController\n",
    "\n",
    "%matplotlib inline\n",
    "plt.rcParams['figure.figsize'] = (10, 15)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "def generate_data(batch_size, length, size):\n",
    "    input_data = np.zeros((batch_size, 2 * length + 2, size), dtype=np.float32)\n",
    "    target_output = np.zeros((batch_size, 2 * length + 2, size), dtype=np.float32)\n",
    "\n",
    "    sequence = np.random.binomial(1, 0.5, (batch_size, length, size - 2))\n",
    "    input_data[:, 0, 0] = 1\n",
    "    input_data[:, 1:length+1, 1:size-1] = sequence\n",
    "    input_data[:, length+1, -1] = 1  # the end symbol\n",
    "    target_output[:, length + 2:, 1:size-1] = sequence\n",
    "\n",
    "    return input_data, target_output\n",
    "\n",
    "def llprint(message):\n",
    "    sys.stdout.write(message)\n",
    "    sys.stdout.flush()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "def binary_cross_entropy(predictions, targets):\n",
    "    return tf.reduce_mean(-1 * targets * tf.log(predictions) - (1 - targets) * tf.log(1 - predictions))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": true,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "def visualize_op(input_series, ntm_output, memory_view):\n",
    "    gs = gridspec.GridSpec(20, 1, hspace=0)\n",
    "\n",
    "    ww_strip = np.squeeze(memory_view['write_weightings'])\n",
    "    rw_strip = np.squeeze(memory_view['read_weightings'])\n",
    "    colored_write = np.zeros((ww_strip.shape[0], ww_strip.shape[1], 3))\n",
    "    colored_read = np.zeros((rw_strip.shape[0], rw_strip.shape[1], 3))\n",
    "    for i in range(ww_strip.shape[0]):\n",
    "        for j in range(ww_strip.shape[1]):\n",
    "            colored_read[i, j] = [rw_strip[i,j], 0., 0.]\n",
    "            colored_write[i, j] = [0., ww_strip[i,j], 0.]\n",
    "\n",
    "\n",
    "    iax = plt.subplot(gs[0:5, 0])\n",
    "    oax = plt.subplot(gs[5:10, 0])\n",
    "    memax = plt.subplot(gs[10:, 0])\n",
    "\n",
    "    iax.grid(True, color='gray')\n",
    "    oax.grid(True, color='gray')\n",
    "    memax.grid(True, color='gray', axis='x')\n",
    "        \n",
    "    iax.imshow(np.squeeze(input_series.T), cmap=plt.cm.gray, interpolation='nearest')\n",
    "    iax.set_ylabel(\"Inputs\")\n",
    "    iax.set_yticks([])\n",
    "\n",
    "    oax.imshow(np.squeeze(ntm_output.T), cmap=plt.cm.gray, interpolation='nearest')\n",
    "    oax.set_ylabel(\"Outputs\")\n",
    "    oax.set_yticks([])\n",
    "\n",
    "    memax.imshow(np.transpose(colored_write + colored_read, [1, 0, 2]), interpolation='nearest')\n",
    "    memax.set_ylabel(\"Memory Location\")\n",
    "    write_legend = mpatches.Rectangle((1,1), 1, 1, color='green', label='Write Head')\n",
    "    read_legend = mpatches.Rectangle((1,1), 1, 1, color='red', label='Read Head')\n",
    "    memax.legend(bbox_to_anchor=(0.21, -0.1), handles=[write_legend, read_legend])\n",
    "    \n",
    "    return colored_write, colored_read"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "def hamming_distance(s1, s2):\n",
    "    \"\"\"Return the Hamming distance between equal-length sequences\"\"\"\n",
    "    if len(s1) != len(s2):\n",
    "        raise ValueError(\"Undefined for sequences of unequal length\")\n",
    "    return sum(el1 != el2 for el1, el2 in zip(s1, s2))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "## Loading and running NTM model\n",
    "\n",
    "Trained on sequences of length up to 10."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Avg. Accuracy: 0.5000\n",
      "Avg. Loss: 0.0047\n",
      "Avg. Dist: 1.2000\n"
     ]
    }
   ],
   "source": [
    "testing_seq_length = 30    # Sequence length for testing\n",
    "testing_runs = 10          # Number of testing runs\n",
    "\n",
    "dist = []\n",
    "losses = []\n",
    "matches = []\n",
    "inputs = []\n",
    "outs = []\n",
    "views = []\n",
    "\n",
    "ckpts_dir = os.path.join(os.path.dirname(\"__file__\"), 'checkpoints')\n",
    "\n",
    "tf.reset_default_graph()\n",
    "\n",
    "with tf.Session() as session:\n",
    "\n",
    "    turing_machine = NTM(\n",
    "        RecurrentController,\n",
    "        input_size = 10,\n",
    "        output_size = 10,\n",
    "        memory_locations = 128,\n",
    "        memory_word_size = 20,\n",
    "        memory_read_heads = 1,\n",
    "        shift_range = 1,\n",
    "        batch_size = 1\n",
    "    )\n",
    "    \n",
    "    outputs, memory_views = turing_machine.get_outputs()\n",
    "    squashed_output = tf.clip_by_value(tf.sigmoid(outputs), 1e-6, 1. - 1e-6)\n",
    "    loss = binary_cross_entropy(squashed_output, turing_machine.target_output)\n",
    "    \n",
    "    session.run(tf.global_variables_initializer())\n",
    "    \n",
    "    # Restoring the provided model\n",
    "    turing_machine.restore(session, ckpts_dir, 'step-100000') \n",
    "    \n",
    "    for i in range(testing_runs):\n",
    "        input_data, target_output = generate_data(1, testing_seq_length, 10)\n",
    "\n",
    "        loss_value,out,mem = session.run([\n",
    "            loss,\n",
    "            squashed_output,\n",
    "            memory_views\n",
    "        ], feed_dict={\n",
    "            turing_machine.input_data: input_data,\n",
    "            turing_machine.target_output: target_output,\n",
    "            turing_machine.sequence_length: testing_seq_length*2 + 2\n",
    "        })\n",
    "        \n",
    "        dist.append(hamming_distance(np.reshape(np.round(out),(1,-1)).tolist()[0],np.reshape(target_output,(1,-1)).tolist()[0]))\n",
    "        losses.append(loss_value)\n",
    "        inputs.append(input_data)\n",
    "        outs.append(out)\n",
    "        views.append(mem)\n",
    "        matches.append(np.allclose(target_output, np.around(out)))\n",
    "        \n",
    "print \"Avg. Accuracy: %.4f\" % (np.mean(matches))\n",
    "print \"Avg. Loss: %.4f\" % (np.mean(losses))\n",
    "print \"Avg. Dist: %.4f\" % (np.mean(dist))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "## Plotting the best output"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Hamming distance: 0\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAk0AAANXCAYAAAAowGsRAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nmgAAABwgNAEAADjwv/4SB1XAsB9WAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fb1cc454350>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "best_indx = np.argmin(dist)\n",
    "print('Hamming distance: %d' % dist[best_indx])\n",
    "best_input, best_output, best_memview = inputs[best_indx], outs[best_indx], views[best_indx]\n",
    "\n",
    "a = visualize_op(best_input, best_output, best_memview)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "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": "2.7.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}