EMF_RBM_Ominglot-20faster.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#  EMF RBM Omniglot\n",
    "\n",
    "### Binarized, 200 Epochs, faster learning rate, with momentum\n",
    "\n",
    "Look at different ways of computing the\n",
    "\n",
    "- energy density\n",
    "- free energy density\n",
    "- entropy density\n",
    "\n",
    "We can\n",
    "\n",
    "- use the TAP S and MF Energy *U_naive   \n",
    "    - which takes an equlibration step first  \n",
    "    - S(mv, mh), U_naive(mv, mh)\n",
    "- use a simple h = means_hiddens() step\n",
    "    - S(v, h), E_MF(v, h)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-03-06T14:43:11.684773",
     "start_time": "2017-03-06T14:43:09.146406"
    },
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/charlesmartin14/anaconda/lib/python2.7/site-packages/matplotlib/font_manager.py:273: UserWarning: Matplotlib is building the font cache using fc-list. This may take a moment.\n",
      "  warnings.warn('Matplotlib is building the font cache using fc-list. This may take a moment.')\n",
      "/Users/charlesmartin14/anaconda/lib/python2.7/site-packages/sklearn/cross_validation.py:44: DeprecationWarning: This module was deprecated in version 0.18 in favor of the model_selection module into which all the refactored classes and functions are moved. Also note that the interface of the new CV iterators are different from that of this module. This module will be removed in 0.20.\n",
      "  \"This module will be removed in 0.20.\", DeprecationWarning)\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "import h5py\n",
    "\n",
    "import matplotlib\n",
    "import matplotlib as mpl\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline \n",
    "\n",
    "from sklearn import linear_model, datasets, metrics, preprocessing \n",
    "from sklearn.cross_validation import train_test_split\n",
    "from sklearn.metrics import accuracy_score\n",
    "\n",
    "import psutil"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-03-06T14:43:11.694119",
     "start_time": "2017-03-06T14:43:11.686403"
    },
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'divide': 'warn', 'invalid': 'warn', 'over': 'warn', 'under': 'ignore'}"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.datasets import fetch_mldata\n",
    "from sklearn.utils.validation import assert_all_finite\n",
    "from scipy.sparse import csc_matrix, csr_matrix, lil_matrix\n",
    "from sklearn.preprocessing import Binarizer\n",
    "np.seterr(all='warn')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-03-06T14:43:12.352957",
     "start_time": "2017-03-06T14:43:12.346312"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from scipy.io import loadmat\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib as mpl\n",
    "\n",
    "def show_image(image): \n",
    "    fig = plt.figure()\n",
    "    ax = fig.add_subplot(1,1,1)\n",
    "    imgplot = ax.imshow(image, cmap=mpl.cm.Greys)\n",
    "    imgplot.set_interpolation('nearest')\n",
    "    ax.xaxis.set_ticks_position('top')\n",
    "    ax.yaxis.set_ticks_position('left')\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-03-06T14:43:12.562232",
     "start_time": "2017-03-06T14:43:12.556901"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import rbm_datasets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-03-06T14:43:15.063167",
     "start_time": "2017-03-06T14:43:12.747815"
    },
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "train_x, train_y, train_char, test_x, test_y, test_char = rbm_datasets.load_omniglot_iwae()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-03-06T14:43:15.261333",
     "start_time": "2017-03-06T14:43:15.064759"
    },
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from sklearn.preprocessing import binarize\n",
    "X_train = binarize(train_x, threshold=0.5, copy=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-03-06T14:43:16.757845",
     "start_time": "2017-03-06T14:43:15.262966"
    },
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP4AAAD7CAYAAABKWyniAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAC29JREFUeJzt3U+opfV9x/H3Jx26SAZUJDrFsWNDoWQjQ0qymS5OKCRS\nCiNZmNSNhhJc1FTaTaybexddtFkIdpFNNGEskSYREjWQ1hQD4iJRtNOMySQplLG1dW6ToOLsQv12\ncZ4Zr9Mz9xzv+Tv3+37Bwec85z73+Z6f93N+v9/zO+dMqgpJvbxv3QVIWj2DLzVk8KWGDL7UkMGX\nGjL4UkMrC36S25L8NMnPk3xhVeedVZJzSf41yb8keX4D6nkkyU6SH+3ad12Sp5P8LMk/Jblmw+rb\nSvJqkpeG221rrO9okmeS/DjJmSR/PuzfiDacUN/nh/0racOsYh0/yfuAnwN/CPw38ALwmar66dJP\nPqMk/w78flW9vu5aAJL8AXABeLSqbh32/S3wq6r64vDieV1V3b9B9W0Bb1XVg+uoabckR4AjVXU6\nyWHgReAk8Fk2oA33qO/TrKANV9Xjfwz4t6p6pap+DfwD4ye5ScIGTX2q6jng8hehk8CpYfsUcPtK\ni9rlCvXBuB3XrqrOV9XpYfsCcBY4yoa04RXqu2l4eOltuKo/9JuA/9x1/1XeeZKbooDvJXkhyefW\nXcwV3FBVOzD+wwFuWHM9k9yb5HSSh9c5FdktyS3AceAHwI2b1oa76vvhsGvpbbgxPdwGOFFVHwH+\nCPizYSi76Tbt/dZfAj5UVceB88AmDPkPA48D9w096+VtttY2nFDfStpwVcH/L+C3d90/OuzbGFX1\n2vDfXwDfYjw92TQ7SW6ES3PE/1lzPe9SVb+ody4afRn46DrrSXKIcaj+vqqeGHZvTBtOqm9Vbbiq\n4L8A/G6SY0l+E/gM8OSKzj1VkvcPr7wk+QDwCeDl9VYFjOd6u+d7TwJ3D9t3AU9cfsCKvau+IUgX\nfYr1t+FXgJ9U1UO79m1SG/6/+lbVhiu5qg/j5TzgIcYvNo9U1d+s5MQzSPI7jHv5Ag4BX1t3fUke\nA0bA9cAOsAV8G/gmcDPwCnBHVb2xQfV9nPFc9W3gHHDPxfn0Guo7ATwLnGH8/7WAB4DngW+w5jbc\no747WUEbriz4kjaHF/ekhgy+1JDBlxoy+FJDcwV/0z94I2myfV/Vn/WDN0lcNpDWpKomvu9/nh5/\n5g/eVNWl29bW1rvub9rN+g5ufZtc2zLq28s8wb8aPngjaQIv7kkNHZrj2Jk/eLO9vX1p+9prr53j\nlMs3Go3WXcKerG//Nrk2WG1981zc+w3gZ4wv7r3G+D3Qf1JVZy/7udrvOSTtXxLqChf39t3jV9X/\nJrkXeJp3PnhzdsphkjbA0j+kY48vrcdePb4X96SGDL7UkMGXGjL4UkMGX2rI4EsNGXypIYMvNWTw\npYYMvtSQwZcaMvhSQwZfasjgSw0ZfKkhgy81ZPClhgy+1JDBlxoy+FJDBl9qyOBLDRl8qSGDLzVk\n8KWGDL7UkMGXGjL4UkMGX2rI4EsNGXypoUPzHJzkHPAm8Dbw66r62CKKkrRccwWfceBHVfX6IoqR\ntBrzDvWzgN8hacXmDW0B30vyQpLPLaIgScs371D/RFW9luSDjF8AzlbVc5f/0Pb29qXt0WjEaDSa\n87SS5pGqWswvSraAt6rqwcv216LOIWl2SaiqTHps30P9JO9PcnjY/gDwCeDl/f4+Saszz1D/RuBb\nSWr4PV+rqqcXU5akZVrYUP+KJ3CoL63FUob6kq5eBl9qyOBLDRl8qSGDLzVk8KWGDL7UkMGXGjL4\nUkMGX2rI4EsNGXypIYMvNWTwpYYMvtTQvN+5pwMumfhx7quG3wUxmT2+1JDBlxoy+FJDBl9qyOBL\nDRl8qSGDLzXkOv6Gcx19b9PaZ9rjXdf57fGlhgy+1JDBlxoy+FJDBl9qyOBLDRl8qaGp6/hJHgH+\nGNipqluHfdcBXweOAeeAO6rqzSXWedVa9jp813Xoi6Y9f9f5J5ulx/8q8MnL9t0P/HNV/R7wDPBX\niy5M0vJMDX5VPQe8ftnuk8CpYfsUcPuC65K0RPud499QVTsAVXUeuGFxJUlatkW9V3/PidD29val\n7dFoxGg0WtBpJe1HZrl4keQY8NSui3tngVFV7SQ5Any/qj58hWProF4gmYUX99Zr3va/mts3CVU1\nsQFmHepnuF30JHD3sH0X8MS+q5O0clN7/CSPASPgemAH2AK+DXwTuBl4hfFy3htXON4ef4k6t+0s\n7PEn9/gzDfXnPPmBDn7nP6yD4CCv4y9iqC/pADH4UkMGX2rI4EsNGXypIYMvNWTwpYb8Xv0pXKfv\n7aCu89vjSw0ZfKkhgy81ZPClhgy+1JDBlxoy+FJD7dfxu6/Td3/+834v/9XKHl9qyOBLDRl8qSGD\nLzVk8KWGDL7UkMGXGmq/jq/eDuo6/TT2+FJDBl9qyOBLDRl8qSGDLzVk8KWGDL7U0NTgJ3kkyU6S\nH+3at5Xk1SQvDbfbllumlqWq5roddAf1+c/S438V+OSE/Q9W1UeG2z8uuC5JSzQ1+FX1HPD6hId6\nvuVJOgDmmePfm+R0koeTXLOwiiQtXWaZpyQ5BjxVVbcO9z8I/LKqKslfA79VVX96hWNra2vr0v3R\naMRoNFpE7QvR/Tvnujuo/zYejJ9bVU18gvsK/qyPDY/XJjeewe+ta/BnHeqHXXP6JEd2PfYp4OX9\nlydp1aZ+LDfJY8AIuD7JfwBbwMeTHAfeBs4B9yyxRkkLNtNQf64TbPhQf5qDPBTsoPNUbhFDfUkH\niMGXGjL4UkMGX2rI4EsNGXypIYMvNeT36s+p8zrxKiz7e++7tr89vtSQwZcaMvhSQwZfasjgSw0Z\nfKkhgy815Dr+FNPWeeddZ+7+PoDuz39d7PGlhgy+1JDBlxoy+FJDBl9qyOBLDRl8qSHX8ee0gn+X\nYK7Hr3au0y+HPb7UkMGXGjL4UkMGX2rI4EsNGXypIYMvNTR1HT/JUeBR4EbgbeDLVfV3Sa4Dvg4c\nA84Bd1TVm0ustSXXsbUMmeGLJo4AR6rqdJLDwIvASeCzwK+q6otJvgBcV1X3Tzi+/OOVVi8JVTXx\nHV5Th/pVdb6qTg/bF4CzwFHG4T81/Ngp4PbFlCtp2d7THD/JLcBx4AfAjVW1A+MXB+CGRRcnaTlm\nfq/+MMx/HLivqi4kuXz8fsXx/Pb29qXt0WjEaDR6b1VKWqipc3yAJIeA7wDfraqHhn1ngVFV7QzX\nAb5fVR+ecKxzfGkN5prjD74C/ORi6AdPAncP23cBT+y7QkkrNctV/RPAs8AZxsP5Ah4Ange+AdwM\nvMJ4Oe+NCcfb40trsFePP9NQf86TG3xpDRYx1Jd0gBh8qSGDLzVk8KWGDL7UkMGXGjL4UkMGX2rI\n4EsNGXypIYMvNWTwpYYMvtSQwZcaMvhSQwZfasjgSw0ZfKkhgy81ZPClhgy+1JDBlxoy+FJDBl9q\nyOBLDRl8qSGDLzVk8KWGDL7UkMGXGpoa/CRHkzyT5MdJziT5/LB/K8mrSV4abrctv1xJi5Bp/3Z9\nkiPAkao6neQw8CJwEvg08FZVPTjl+Jp2DkmLl4SqyqTHDk07uKrOA+eH7QtJzgI3XfzdC6tS0sq8\npzl+kluA48APh133Jjmd5OEk1yy4NklLMnPwh2H+48B9VXUB+BLwoao6znhEsOeQX9LmmDrHB0hy\nCPgO8N2qemjC48eAp6rq1gmP1dbW1qX7o9GI0Wg0T82SZrDXHH/W4D8K/LKq/nLXviPD/J8kfwF8\ntKrunHCsF/ekNZgr+ElOAM8CZ4Aabg8AdzKe778NnAPuqaqdCccbfGkN5u7x5zy5wZfWYK/g+849\nqSGDLzVk8KWGDL7UkMGXGjL4UkMGX2rI4EsNGXypIYMvNWTwpYYMvtSQwZcaMvhSQwZfasjgSw1N\n/XrtRUj8Fm5pkyz9G3gkbR6H+lJDBl9qyOBLDRl8qSGDLzX0fyPSo2uOeHoJAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x10df43d10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP4AAAD7CAYAAABKWyniAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n3GJFAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x10df98510>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP4AAAD7CAYAAABKWyniAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nqaH/A8HLb+A3rtIuAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11567cbd0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP4AAAD7CAYAAABKWyniAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nyrhgAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1158fbb50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP4AAAD7CAYAAABKWyniAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nkbR5HOpLDRl8qSGDLzVk8KWGDL7U0P8Bkxn2bqQ6IxMAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1159a3dd0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP4AAAD7CAYAAABKWyniAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x115cd0f50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP4AAAD7CAYAAABKWyniAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nSZvHob7UkMGXGjL4UkMGX2rI4EsN/R9vdr4SGWQSJgAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x115e2e3d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP4AAAD7CAYAAABKWyniAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nA8M1CYi88ONlAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x115ecbc50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP4AAAD7CAYAAABKWyniAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1160d4810>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP4AAAD7CAYAAABKWyniAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nkBDBBxL6f/73tRBGPkMfAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x115d6cf90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for i in range(5):\n",
    "    show_image(X_train[i+100].reshape(28,28))\n",
    "    show_image(train_x[i+100].reshape(28,28))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-03-06T14:43:16.762767",
     "start_time": "2017-03-06T14:43:16.759412"
    },
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(24345, 784)"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2016-10-20T17:56:14.156132",
     "start_time": "2016-10-20T17:56:14.153682"
    }
   },
   "source": [
    "### EMF RBM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-03-06T14:43:27.827304",
     "start_time": "2017-03-06T14:43:27.822781"
    },
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "sswap(total=9663676416L, used=7910719488L, free=1752956928L, percent=81.9, sin=64430620672L, sout=624091136L)"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "psutil.swap_memory()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### mean Entropy, Energy, and Free Energy\n",
    "\n",
    "need to grid search these parameters and plot"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-03-06T14:53:32.326715",
     "start_time": "2017-03-06T14:48:07.569926"
    },
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EMF_RBM(batch_size=100, decay=0.1, learning_rate=0.1, momentum=0.25,\n",
      "    monitor=True, n_components=256, n_iter=20, neq_steps=3,\n",
      "    random_state=<mtrand.RandomState object at 0x106e7a3c0>, sigma=0.01,\n",
      "    thresh=1e-08, verbose=True, weight_decay='L1')\n",
      "[EMF_RBM] Iteration 1, pseudo-likelihood = -148.13, time = 6.34s\n",
      "computing TAP Free Energies\n",
      "monitor:  -181.827475361 286.062118642 106.880495497\n",
      "[EMF_RBM] Iteration 2, pseudo-likelihood = -129.24, time = 15.35s\n",
      "computing TAP Free Energies\n",
      "monitor:  -177.063406042 257.423971435 86.6211306868\n",
      "[EMF_RBM] Iteration 3, pseudo-likelihood = -115.13, time = 15.73s\n",
      "computing TAP Free Energies\n",
      "monitor:  -185.663694989 244.436920555 68.3220989913\n",
      "[EMF_RBM] Iteration 4, pseudo-likelihood = -104.67, time = 18.72s\n",
      "computing TAP Free Energies\n",
      "monitor:  -197.70002331 235.179934571 49.6281101364\n",
      "[EMF_RBM] Iteration 5, pseudo-likelihood = -100.64, time = 16.70s\n",
      "computing TAP Free Energies\n",
      "monitor:  -211.173697537 228.738151639 31.7645808548\n",
      "[EMF_RBM] Iteration 6, pseudo-likelihood = -94.97, time = 16.16s\n",
      "computing TAP Free Energies\n",
      "monitor:  -224.582977475 223.541437661 14.8289481977\n",
      "[EMF_RBM] Iteration 7, pseudo-likelihood = -91.23, time = 15.92s\n",
      "computing TAP Free Energies\n",
      "monitor:  -236.937379149 219.02822565 -0.668928099905\n",
      "[EMF_RBM] Iteration 8, pseudo-likelihood = -92.14, time = 16.07s\n",
      "computing TAP Free Energies\n",
      "monitor:  -247.866941353 215.094541214 -14.3663411822\n",
      "[EMF_RBM] Iteration 9, pseudo-likelihood = -84.69, time = 15.83s\n",
      "computing TAP Free Energies\n",
      "monitor:  -257.412556754 211.647229817 -26.3370779937\n",
      "[EMF_RBM] Iteration 10, pseudo-likelihood = -83.23, time = 15.77s\n",
      "computing TAP Free Energies\n",
      "monitor:  -265.675050073 208.56465564 -36.7723442544\n",
      "[EMF_RBM] Iteration 11, pseudo-likelihood = -81.27, time = 16.05s\n",
      "computing TAP Free Energies\n",
      "monitor:  -272.81086998 205.791623188 -45.8589437013\n",
      "[EMF_RBM] Iteration 12, pseudo-likelihood = -81.56, time = 15.77s\n",
      "computing TAP Free Energies\n",
      "monitor:  -278.974974192 203.249711964 -53.8219844072\n",
      "[EMF_RBM] Iteration 13, pseudo-likelihood = -80.37, time = 16.25s\n",
      "computing TAP Free Energies\n",
      "monitor:  -284.327689599 200.930468249 -60.8162517571\n",
      "[EMF_RBM] Iteration 14, pseudo-likelihood = -77.13, time = 16.26s\n",
      "computing TAP Free Energies\n",
      "monitor:  -289.017645824 198.829412175 -66.9797877326\n",
      "[EMF_RBM] Iteration 15, pseudo-likelihood = -76.72, time = 18.19s\n",
      "computing TAP Free Energies\n",
      "monitor:  -293.186176613 196.941727802 -72.4457207366\n",
      "[EMF_RBM] Iteration 16, pseudo-likelihood = -74.92, time = 16.13s\n",
      "computing TAP Free Energies\n",
      "monitor:  -296.909062167 195.243053839 -77.3139575428\n",
      "[EMF_RBM] Iteration 17, pseudo-likelihood = -73.56, time = 16.79s\n",
      "computing TAP Free Energies\n",
      "monitor:  -300.273937099 193.698458246 -81.707030137\n",
      "[EMF_RBM] Iteration 18, pseudo-likelihood = -73.94, time = 16.06s\n",
      "computing TAP Free Energies\n",
      "monitor:  -303.329809811 192.288912228 -85.6888180207\n",
      "[EMF_RBM] Iteration 19, pseudo-likelihood = -73.46, time = 15.76s\n",
      "computing TAP Free Energies\n",
      "monitor:  -306.12901152 190.998302646 -89.3272470662\n",
      "[EMF_RBM] Iteration 20, pseudo-likelihood = -68.45, time = 15.59s\n",
      "computing TAP Free Energies\n",
      "monitor:  -308.708658206 189.81245044 -92.6708123546\n"
     ]
    }
   ],
   "source": [
    "import emf_rbm\n",
    "rbm = emf_rbm.EMF_RBM(verbose=True, monitor=True, learning_rate=0.1, n_iter=20,momentum=0.25, sigma=0.01, decay=0.1)\n",
    "print rbm\n",
    "rbm= rbm.fit(X_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-03-06T14:57:10.470643",
     "start_time": "2017-03-06T14:57:10.181457"
    },
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.text.Text at 0x10edb20d0>"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZYAAAEKCAYAAAAxXHOuAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nMzNrKTcWMzNrKTcWMzNrKTcWMzNrqT9ZhghSvtppuAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x115eea750>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(rbm.entropies-rbm.entropies[0], color='red', label='Entropy')\n",
    "plt.plot(rbm.mean_field_energies-rbm.mean_field_energies[0], color='blue', label='MF Energy')\n",
    "plt.plot(rbm.free_energies-rbm.free_energies[0], color='green', label='Free Energy')\n",
    "plt.legend(loc='upper right')\n",
    "plt.title(\"Ominglot: Free Energy, Mean Field Enegry, & Entropy vs Epochs\")\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Entropy does not flatline so fast"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-03-06T14:57:14.700369",
     "start_time": "2017-03-06T14:57:14.591233"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.preprocessing import binarize\n",
    "X_test = binarize(test_x, threshold=0.5, copy=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-03-06T14:57:14.765628",
     "start_time": "2017-03-06T14:57:14.759867"
    },
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16,\n",
       "       17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33,\n",
       "       34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49])"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.unique(train_y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-03-06T15:46:19.939087",
     "start_time": "2017-03-06T15:15:29.520125"
    },
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LinearSVC(C=0.01, class_weight=None, dual=True, fit_intercept=True,\n",
      "     intercept_scaling=1, loss='squared_hinge', max_iter=1000,\n",
      "     multi_class='ovr', penalty='l2', random_state=None, tol=0.0001,\n",
      "     verbose=0) 0.256133828996\n",
      "LinearSVC(C=0.1, class_weight=None, dual=True, fit_intercept=True,\n",
      "     intercept_scaling=1, loss='squared_hinge', max_iter=1000,\n",
      "     multi_class='ovr', penalty='l2', random_state=None, tol=0.0001,\n",
      "     verbose=0) 0.260346964064\n",
      "LinearSVC(C=1, class_weight=None, dual=True, fit_intercept=True,\n",
      "     intercept_scaling=1, loss='squared_hinge', max_iter=1000,\n",
      "     multi_class='ovr', penalty='l2', random_state=None, tol=0.0001,\n",
      "     verbose=0) 0.26009913259\n",
      "LinearSVC(C=10, class_weight=None, dual=True, fit_intercept=True,\n",
      "     intercept_scaling=1, loss='squared_hinge', max_iter=1000,\n",
      "     multi_class='ovr', penalty='l2', random_state=None, tol=0.0001,\n",
      "     verbose=0) 0.212639405204\n",
      "LinearSVC(C=100, class_weight=None, dual=True, fit_intercept=True,\n",
      "     intercept_scaling=1, loss='squared_hinge', max_iter=1000,\n",
      "     multi_class='ovr', penalty='l2', random_state=None, tol=0.0001,\n",
      "     verbose=0) 0.125526641884\n"
     ]
    }
   ],
   "source": [
    "from sklearn.svm import LinearSVC\n",
    "from sklearn.metrics import accuracy_score\n",
    "\n",
    "F_train = rbm.transform(X_train)\n",
    "F_test = rbm.transform(X_test)\n",
    "\n",
    "for c in [0.01, 0.1, 1, 10, 100]:\n",
    "    classifier = LinearSVC(C=c)\n",
    "    classifier.fit(F_train, train_y)\n",
    "    test_emf_rbm_pred_y = classifier.predict(F_test)\n",
    "    emf_accuracy = accuracy_score(y_pred=test_emf_rbm_pred_y, y_true=test_y)\n",
    "    print classifier, emf_accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-03-05T23:13:15.864513",
     "start_time": "2017-03-06T07:06:50.647Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python [Root]",
   "language": "python",
   "name": "Python [Root]"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}