ATM experiment

atm_experiment.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from atm import ATM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 199,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# データ生成\n",
    "vocab = {'computer':0, 'banana':1, 'apple':2}\n",
    "authors = {'Yamada':0, 'Suzuki':1}\n",
    "\n",
    "V = len(vocab)\n",
    "S = len(authors)\n",
    "\n",
    "W = [] # words\n",
    "A = [] # authors\n",
    "\n",
    "W.append(map(lambda v:vocab[v], ['computer', 'computer']))\n",
    "A.append(map(lambda a:authors[a], ['Yamada']))\n",
    "W.append(map(lambda v:vocab[v], ['banana', 'banana']))\n",
    "A.append(map(lambda a:authors[a], ['Suzuki']))\n",
    "W.append(map(lambda v:vocab[v], ['computer', 'computer', 'apple']))\n",
    "A.append(map(lambda a:authors[a], ['Yamada']))\n",
    "W.append(map(lambda v:vocab[v], ['banana', 'banana', 'apple']))\n",
    "A.append(map(lambda a:authors[a], ['Suzuki']))\n",
    "W.append(map(lambda v:vocab[v], ['apple']))\n",
    "A.append(map(lambda a:authors[a], ['Yamada']))\n",
    "W.append(map(lambda v:vocab[v], ['apple']))\n",
    "A.append(map(lambda a:authors[a], ['Suzuki']))\n",
    "W.append(map(lambda v:vocab[v], ['computer', 'computer']))\n",
    "A.append(map(lambda a:authors[a], ['Suzuki', 'Yamada']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 200,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# モデル定義\n",
    "K = 2\n",
    "alpha = 0.01\n",
    "beta= 0.01\n",
    "max_iter = 1000\n",
    "\n",
    "atm = ATM(K=K, alpha=alpha, beta=beta, max_iter=max_iter)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 201,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<atm.ATM instance at 0x10e8a0638>"
      ]
     },
     "execution_count": 201,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Author Topic Model のフィッティング\n",
    "atm.fit(W,A,V,S)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 202,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[array([0, 0]), array([1, 1]), array([0, 0, 0]), array([1, 1, 1]), array([0]), array([1]), array([0, 0])]\n",
      "[array([0, 0]), array([0, 0]), array([0, 0, 0]), array([0, 0, 0]), array([0]), array([0]), array([1, 1])]\n"
     ]
    }
   ],
   "source": [
    "# 正しくトピックが割り当てられている\n",
    "# ５番目の文書の \"apple\" はコンピューターの話題ばかり書く山田に書かれたので、コンピュータトピックが割り当てられている\n",
    "# ６番目の文書の \"apple\" は食べ物の話題ばかり書く鈴木に書かれたので、食べ物トピックが割り当てられている\n",
    "# ７番目の文書は二人の共著になっているが、鈴木は食べ物の話題しか書けないので、明らかに山田だけで書いたことが分かる\n",
    "print atm.max_Z\n",
    "print atm.max_Y"
   ]
  }
 ],
 "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
}