CTM experiments

ctm_experiment.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from jtm import JTM # Joint Topic Model\n",
    "from ctm import CTM # Correspondence Topic Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# データ生成\n",
    "vocab = {'computer':0, 'banana':1, 'ipad':2, 'orange':3, 'windows':4}\n",
    "categories = {'FOODS':0, 'TECH':1}\n",
    "\n",
    "Vw = len(vocab)\n",
    "Vx = len(categories)\n",
    "\n",
    "W = [] # words\n",
    "X = [] # categories\n",
    "\n",
    "W.append(map(lambda v:vocab[v], ['computer', 'ipad', 'ipad']))\n",
    "X.append(map(lambda c:categories[c], ['TECH']))\n",
    "W.append(map(lambda v:vocab[v], ['ipad', 'ipad']))\n",
    "X.append(map(lambda c:categories[c], ['TECH']))\n",
    "W.append(map(lambda v:vocab[v], ['ipad', 'ipad']))\n",
    "X.append(map(lambda c:categories[c], ['TECH']))\n",
    "W.append(map(lambda v:vocab[v], ['ipad', 'ipad']))\n",
    "X.append(map(lambda c:categories[c], ['TECH']))\n",
    "W.append(map(lambda v:vocab[v], ['banana', 'orange']))\n",
    "X.append(map(lambda c:categories[c], ['FOODS']))\n",
    "W.append(map(lambda v:vocab[v], ['banana', 'orange']))\n",
    "X.append(map(lambda c:categories[c], ['FOODS']))\n",
    "W.append(map(lambda v:vocab[v], ['windows', 'windows']))\n",
    "X.append(map(lambda c:categories[c], ['TECH']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# モデル定義\n",
    "K = 3\n",
    "alpha = 0.1\n",
    "beta= 0.1\n",
    "gamma= 0.1\n",
    "max_iter = 100\n",
    "\n",
    "jtm = JTM(K=K, alpha=alpha, beta=beta, max_iter=max_iter)\n",
    "ctm = CTM(K=K, alpha=alpha, beta=beta, gamma=gamma, max_iter=max_iter)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<ctm.CTM instance at 0x10a7a7fc8>"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# CTMのフィッティング\n",
    "ctm.fit(W,X,Vw,Vx)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[array([2, 2, 2]), array([2, 2]), array([2, 2]), array([2, 2]), array([1, 1]), array([1, 1]), array([0, 0])]\n",
      "[array([2]), array([2]), array([2]), array([2]), array([1]), array([1]), array([0])]\n"
     ]
    }
   ],
   "source": [
    "# うまいことトピック割当できているし、\n",
    "# 単語に割り当てられていないトピックが付加情報に割り当てられることもない\n",
    "print ctm.Z\n",
    "print ctm.Y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<jtm.JTM instance at 0x10a7f1098>"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# JTMのフィティング\n",
    "jtm.fit([W,X],[Vw,Vx])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[array([1, 0, 0]), array([0, 0]), array([0, 0]), array([0, 0]), array([2, 2]), array([2, 2]), array([1, 1])]\n",
      "[array([0]), array([0]), array([0]), array([0]), array([2]), array([2]), array([0])]\n"
     ]
    }
   ],
   "source": [
    "# あまりうまくトピック割り当てができていない\n",
    "# 最後の文書ではトピック 0 は単語に割り当てられていないのに付加情報に割り当てられてしまっている\n",
    "print jtm.Z[0]\n",
    "print jtm.Z[1]"
   ]
  }
 ],
 "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
}