rsa_test.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "%load_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from learning_rsa import LSLTrainer, FeatureVectorizer\n",
    "from collections import defaultdict"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "featurizations_tensor.shape:\n",
      "(4, (3, 6))\n",
      "list[<class 'scipy.sparse.coo.coo_matrix'>]\n",
      "(4, 3, 6)\n",
      "(0,)\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWYAAAC/CAYAAADNYHZFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fc9b7996bd0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1,)\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWYAAAC/CAYAAADNYHZFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\npUO5jfgAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fc9b77e8d10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(2,)\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWYAAAC/CAYAAADNYHZFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n5RwzwDDCDDCMMAMMI8wAwwgzwDA/B/1q5TMnfd3xAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fc9b76f6490>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(3,)\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWYAAAC/CAYAAADNYHZFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\na9h7tbmvSQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fc9b76a3510>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(0, 0, 0) 0.0\n",
      "(0, 0, 1) 0.0\n",
      "(0, 0, 2) 0.0\n",
      "(0, 0, 3) 0.0\n",
      "(0, 0, 4) 0.0\n",
      "(0, 0, 5) 0.0\n",
      "(0, 1, 0) 0.0\n",
      "(0, 1, 1) 0.0\n",
      "(0, 1, 2) 0.0\n",
      "(0, 1, 3) 0.0\n",
      "(0, 1, 4) 0.0\n",
      "(0, 1, 5) 0.0\n",
      "(0, 2, 0) 0.0\n",
      "(0, 2, 1) 0.0\n",
      "(0, 2, 2) 0.0\n",
      "(0, 2, 3) 0.0\n",
      "(0, 2, 4) 0.0\n",
      "(0, 2, 5) 0.0\n",
      "(1, 0, 0) 1.0\n",
      "(1, 0, 1) 1.0\n",
      "(1, 0, 2) 0.0\n",
      "(1, 0, 3) 0.0\n",
      "(1, 0, 4) 0.0\n",
      "(1, 0, 5) 0.0\n",
      "(1, 1, 0) 1.0\n",
      "(1, 1, 1) 0.0\n",
      "(1, 1, 2) 0.0\n",
      "(1, 1, 3) 0.0\n",
      "(1, 1, 4) 1.0\n",
      "(1, 1, 5) 0.0\n",
      "(1, 2, 0) 1.0\n",
      "(1, 2, 1) 0.0\n",
      "(1, 2, 2) 0.0\n",
      "(1, 2, 3) 0.0\n",
      "(1, 2, 4) 0.0\n",
      "(1, 2, 5) 0.0\n",
      "(2, 0, 0) 0.0\n",
      "(2, 0, 1) 0.0\n",
      "(2, 0, 2) 1.0\n",
      "(2, 0, 3) 1.0\n",
      "(2, 0, 4) 0.0\n",
      "(2, 0, 5) 0.0\n",
      "(2, 1, 0) 0.0\n",
      "(2, 1, 1) 0.0\n",
      "(2, 1, 2) 0.0\n",
      "(2, 1, 3) 1.0\n",
      "(2, 1, 4) 0.0\n",
      "(2, 1, 5) 1.0\n",
      "(2, 2, 0) 0.0\n",
      "(2, 2, 1) 0.0\n",
      "(2, 2, 2) 0.0\n",
      "(2, 2, 3) 1.0\n",
      "(2, 2, 4) 0.0\n",
      "(2, 2, 5) 0.0\n",
      "(3, 0, 0) 1.0\n",
      "(3, 0, 1) 1.0\n",
      "(3, 0, 2) 1.0\n",
      "(3, 0, 3) 1.0\n",
      "(3, 0, 4) 0.0\n",
      "(3, 0, 5) 0.0\n",
      "(3, 1, 0) 1.0\n",
      "(3, 1, 1) 0.0\n",
      "(3, 1, 2) 0.0\n",
      "(3, 1, 3) 1.0\n",
      "(3, 1, 4) 1.0\n",
      "(3, 1, 5) 1.0\n",
      "(3, 2, 0) 1.0\n",
      "(3, 2, 1) 0.0\n",
      "(3, 2, 2) 0.0\n",
      "(3, 2, 3) 1.0\n",
      "(3, 2, 4) 0.0\n",
      "(3, 2, 5) 0.0\n",
      "weights.shape:\n",
      "(6, 1)\n",
      "<class 'scipy.sparse.coo.coo_matrix'>\n",
      "(6,)\n",
      "Literal scores:\n",
      "<type 'numpy.ndarray'>\n",
      "(3, 4)\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAUAAAAD2CAYAAACndF52AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fc9b75d1690>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(0, 0) -1.38629436112\n",
      "(0, 1) -1.38629436112\n",
      "(0, 2) -1.38629436112\n",
      "(0, 3) -1.38629436112\n",
      "(1, 0) -1.38629436112\n",
      "(1, 1) -1.38629436112\n",
      "(1, 2) -1.38629436112\n",
      "(1, 3) -1.38629436112\n",
      "(2, 0) -1.38629436112\n",
      "(2, 1) -1.38629436112\n",
      "(2, 2) -1.38629436112\n",
      "(2, 3) -1.38629436112\n",
      "\n",
      "Literal-expected phi:\n",
      "<class 'scipy.sparse.csr.csr_matrix'>\n",
      "(3, 6)\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWYAAAC/CAYAAADNYHZFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\na9h7tbmvSQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fc9b7518490>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(0, 0) 0.5\n",
      "(0, 1) 0.5\n",
      "(0, 2) 0.5\n",
      "(0, 3) 0.5\n",
      "(0, 4) 0.0\n",
      "(0, 5) 0.0\n",
      "(1, 0) 0.5\n",
      "(1, 1) 0.0\n",
      "(1, 2) 0.0\n",
      "(1, 3) 0.5\n",
      "(1, 4) 0.5\n",
      "(1, 5) 0.5\n",
      "(2, 0) 0.5\n",
      "(2, 1) 0.0\n",
      "(2, 2) 0.0\n",
      "(2, 3) 0.5\n",
      "(2, 4) 0.0\n",
      "(2, 5) 0.0\n",
      "\n",
      "Gradient of literal:\n",
      "list[<class 'scipy.sparse.csr.csr_matrix'>]\n",
      "(4, 3, 6)\n",
      "(0,)\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWYAAAC/CAYAAADNYHZFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nDCPMAMMIM8AwwgwwjDADDPN/KD7s+OFszKEAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fc9b75184d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1,)\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWYAAAC/CAYAAADNYHZFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fc9b73ba1d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(2,)\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWYAAAC/CAYAAADNYHZFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fc9b749be10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(3,)\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWYAAAC/CAYAAADNYHZFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\na9h7tbmvSQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fc9b7571650>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(0, 0, 0) -0.5\n",
      "(0, 0, 1) -0.5\n",
      "(0, 0, 2) -0.5\n",
      "(0, 0, 3) -0.5\n",
      "(0, 0, 4) 0.0\n",
      "(0, 0, 5) 0.0\n",
      "(0, 1, 0) -0.5\n",
      "(0, 1, 1) 0.0\n",
      "(0, 1, 2) 0.0\n",
      "(0, 1, 3) -0.5\n",
      "(0, 1, 4) -0.5\n",
      "(0, 1, 5) -0.5\n",
      "(0, 2, 0) -0.5\n",
      "(0, 2, 1) 0.0\n",
      "(0, 2, 2) 0.0\n",
      "(0, 2, 3) -0.5\n",
      "(0, 2, 4) 0.0\n",
      "(0, 2, 5) 0.0\n",
      "(1, 0, 0) 0.5\n",
      "(1, 0, 1) 0.5\n",
      "(1, 0, 2) -0.5\n",
      "(1, 0, 3) -0.5\n",
      "(1, 0, 4) 0.0\n",
      "(1, 0, 5) 0.0\n",
      "(1, 1, 0) 0.5\n",
      "(1, 1, 1) 0.0\n",
      "(1, 1, 2) 0.0\n",
      "(1, 1, 3) -0.5\n",
      "(1, 1, 4) 0.5\n",
      "(1, 1, 5) -0.5\n",
      "(1, 2, 0) 0.5\n",
      "(1, 2, 1) 0.0\n",
      "(1, 2, 2) 0.0\n",
      "(1, 2, 3) -0.5\n",
      "(1, 2, 4) 0.0\n",
      "(1, 2, 5) 0.0\n",
      "(2, 0, 0) -0.5\n",
      "(2, 0, 1) -0.5\n",
      "(2, 0, 2) 0.5\n",
      "(2, 0, 3) 0.5\n",
      "(2, 0, 4) 0.0\n",
      "(2, 0, 5) 0.0\n",
      "(2, 1, 0) -0.5\n",
      "(2, 1, 1) 0.0\n",
      "(2, 1, 2) 0.0\n",
      "(2, 1, 3) 0.5\n",
      "(2, 1, 4) -0.5\n",
      "(2, 1, 5) 0.5\n",
      "(2, 2, 0) -0.5\n",
      "(2, 2, 1) 0.0\n",
      "(2, 2, 2) 0.0\n",
      "(2, 2, 3) 0.5\n",
      "(2, 2, 4) 0.0\n",
      "(2, 2, 5) 0.0\n",
      "(3, 0, 0) 0.5\n",
      "(3, 0, 1) 0.5\n",
      "(3, 0, 2) 0.5\n",
      "(3, 0, 3) 0.5\n",
      "(3, 0, 4) 0.0\n",
      "(3, 0, 5) 0.0\n",
      "(3, 1, 0) 0.5\n",
      "(3, 1, 1) 0.0\n",
      "(3, 1, 2) 0.0\n",
      "(3, 1, 3) 0.5\n",
      "(3, 1, 4) 0.5\n",
      "(3, 1, 5) 0.5\n",
      "(3, 2, 0) 0.5\n",
      "(3, 2, 1) 0.0\n",
      "(3, 2, 2) 0.0\n",
      "(3, 2, 3) 0.5\n",
      "(3, 2, 4) 0.0\n",
      "(3, 2, 5) 0.0\n",
      "\n",
      "Speaker scores:\n",
      "<type 'numpy.ndarray'>\n",
      "(3, 4)\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAUAAAAD2CAYAAACndF52AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fc9b78627d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(0, 0) -1.09861228867\n",
      "(0, 1) -1.09861228867\n",
      "(0, 2) -1.09861228867\n",
      "(0, 3) -1.09861228867\n",
      "(1, 0) -1.09861228867\n",
      "(1, 1) -1.09861228867\n",
      "(1, 2) -1.09861228867\n",
      "(1, 3) -1.09861228867\n",
      "(2, 0) -1.09861228867\n",
      "(2, 1) -1.09861228867\n",
      "(2, 2) -1.09861228867\n",
      "(2, 3) -1.09861228867\n",
      "\n",
      "Speaker-expected gradient of literal:\n",
      "<type 'numpy.ndarray'>\n",
      "(4, 6)\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWYAAAD2CAYAAADlNy3hAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nLAPoePsAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fc9b748ab50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(0, 0) -0.5\n",
      "(0, 1) -0.166666666667\n",
      "(0, 2) -0.166666666667\n",
      "(0, 3) -0.5\n",
      "(0, 4) -0.166666666667\n",
      "(0, 5) -0.166666666667\n",
      "(1, 0) 0.5\n",
      "(1, 1) 0.166666666667\n",
      "(1, 2) -0.166666666667\n",
      "(1, 3) -0.5\n",
      "(1, 4) 0.166666666667\n",
      "(1, 5) -0.166666666667\n",
      "(2, 0) -0.5\n",
      "(2, 1) -0.166666666667\n",
      "(2, 2) 0.166666666667\n",
      "(2, 3) 0.5\n",
      "(2, 4) -0.166666666667\n",
      "(2, 5) 0.166666666667\n",
      "(3, 0) 0.5\n",
      "(3, 1) 0.166666666667\n",
      "(3, 2) 0.166666666667\n",
      "(3, 3) 0.5\n",
      "(3, 4) 0.166666666667\n",
      "(3, 5) 0.166666666667\n",
      "\n",
      "Gradient of speaker:\n",
      "<type 'numpy.ndarray'>\n",
      "(4, 6)\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWYAAAD2CAYAAADlNy3hAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nhBlgGGEGGEaYAYb5P69gLZK+CxmqAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fc9b736dc90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(0, 0) 0.0\n",
      "(0, 1) -0.333333333333\n",
      "(0, 2) -0.333333333333\n",
      "(0, 3) 0.0\n",
      "(0, 4) 0.166666666667\n",
      "(0, 5) 0.166666666667\n",
      "(1, 0) 0.0\n",
      "(1, 1) 0.333333333333\n",
      "(1, 2) -0.333333333333\n",
      "(1, 3) 0.0\n",
      "(1, 4) -0.166666666667\n",
      "(1, 5) 0.166666666667\n",
      "(2, 0) 0.0\n",
      "(2, 1) -0.333333333333\n",
      "(2, 2) 0.333333333333\n",
      "(2, 3) 0.0\n",
      "(2, 4) 0.166666666667\n",
      "(2, 5) -0.166666666667\n",
      "(3, 0) 0.0\n",
      "(3, 1) 0.333333333333\n",
      "(3, 2) 0.333333333333\n",
      "(3, 3) 0.0\n",
      "(3, 4) -0.166666666667\n",
      "(3, 5) -0.166666666667\n",
      "\n",
      "Listener scores:\n",
      "<type 'numpy.ndarray'>\n",
      "(4,)\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWYAAABrCAYAAABEzyh8AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fc9b7773d50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(0,) -1.38629436112\n",
      "(1,) -1.38629436112\n",
      "(2,) -1.38629436112\n",
      "(3,) -1.38629436112\n",
      "\n",
      "Listener-expected gradient of speaker:\n",
      "<type 'numpy.ndarray'>\n",
      "(6,)\n",
      "\n",
      "Gradient of listener (final gradient):\n",
      "<type 'numpy.ndarray'>\n",
      "(6,)\n",
      "0.333333333333 world->bob\n",
      "-0.166666666667 alice->alice\n",
      "-0.166666666667 alice->bob\n",
      "0.333333333333 world->alice\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "defaultdict(<type 'float'>, {'world->bob': 0.33333333333333337, 'alice->alice': -0.16666666666666666, 'alice->bob': -0.16666666666666666, 'world->alice': 0.33333333333333337})"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "def phi_unigrams(phrase, answer):\n",
    "    feats_dict = {'%s->%s' % (w, r): 1.0\n",
    "                  for w in phrase.split() for r in answer}\n",
    "    return defaultdict(float, feats_dict)\n",
    "\n",
    "MESSAGES = ['hello world', 'hello alice', 'hello']\n",
    "REFERENTS = [(), ('alice',), ('bob',), ('alice', 'bob')]\n",
    "\n",
    "'''\n",
    "weights = defaultdict(float,\n",
    "                      {'world->bob': 0.390,\n",
    "                       'world->alice': 0.455,\n",
    "                       'alice->bob': -1.854,\n",
    "                       'alice->alice': 1.315})\n",
    "'''\n",
    "weights = defaultdict(float)\n",
    "\n",
    "vec = FeatureVectorizer(phi_unigrams)\n",
    "for x in MESSAGES:\n",
    "    for y in REFERENTS:\n",
    "        for name in phi_unigrams(x, y):\n",
    "            vec.get_dim(name)\n",
    "\n",
    "trainer = LSLTrainer(phi=phi_unigrams)\n",
    "trainer.vectorizer = vec\n",
    "trainer.gradient('hello world', MESSAGES,\n",
    "                 ('alice', 'bob'), REFERENTS,\n",
    "                 weights,\n",
    "                 verbose=3)"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "Starting from filled-in weights:\n",
    "defaultdict(<type 'float'>, {'world->bob': 0.20945775283539719, 'alice->alice': -0.061718087477370787, 'hello->bob': 0.033327999737709113, 'hello->alice': -0.0067952062066051261, 'alice->bob': -0.086688579560132692, 'world->alice': 0.2683766257299684})\n",
    "\n",
    "Starting from zero weights:\n",
    "defaultdict(<type 'float'>, {'world->bob': 0.33333333333333337, 'alice->alice': -0.16666666666666666, 'alice->bob': -0.16666666666666666, 'world->alice': 0.33333333333333337})\n",
    "\n",
    "3,4,6\n",
    "3,4\n",
    "3,6\n",
    "3,4,6\n",
    "3,4\n",
    "4,6\n",
    "4,6\n",
    "4\n",
    "6\n",
    "6"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "('alice', 'bob')"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "trainer.predict(x='hello',\n",
    "                w=weights,\n",
    "                messages=MESSAGES,\n",
    "                distractors=REFERENTS)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "('alice',)"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from learning import LiteralTrainer\n",
    "LiteralTrainer(phi=phi_unigrams).predict(x='hello',\n",
    "                                         w=weights,\n",
    "                                         distractors=REFERENTS)"
   ]
  }
 ],
 "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": "2.7.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}