tf_factorization_machine.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 250,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import sys\n",
    "import pandas as pd\n",
    "import numpy  as np\n",
    "import tensorflow as tf\n",
    "import matplotlib.pyplot as plt\n",
    "from tqdm import tqdm\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 251,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The autoreload extension is already loaded. To reload it, use:\n",
      "  %reload_ext autoreload\n"
     ]
    }
   ],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "$$ \\sum_i \\sum_j x_{in} \\, x_{jn} \\sum_f \\, V_{fi} \\, V_{fj}   $$\n",
    "$$ \\sum_i \\sum_j x_{in} \\, x_{jn} \\, M_{ij}   $$"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 252,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from tflearn import tflearn, vardict\n",
    "\n",
    "class tffactmachine(tflearn):\n",
    "    def _create_network(self):\n",
    "        \n",
    "        self.vars = vardict()\n",
    "        self.vars[\"x\"] = tf.placeholder(\"float\", shape=[None, self.xlen])\n",
    "        self.vars[\"y\"] = tf.placeholder(\"float\", shape=[None, 1])\n",
    "                    \n",
    "        # Create Model        \n",
    "        with tf.variable_scope('fm') as scope:\n",
    "            \n",
    "            self.parameters[\"bias\"] = tf.Variable(tf.constant(0.1, shape=[1, 1]), name=\"bias\")\n",
    "            self.parameters[\"weights\"] = tf.Variable(tf.truncated_normal([1, self.xlen], stddev=0.01), name=\"weights\")\n",
    "            self.parameters[\"veights\"] = tf.Variable( tf.random_normal([ self.xlen, self.num_factors], stddev=0.005),\n",
    "                                                      name=\"veights\")\n",
    "            \n",
    "            VVT =   np.diag(2*np.ones( self.xlen)) * \\\n",
    "                tf.matmul( self.parameters[\"veights\"], tf.transpose(self.parameters[\"veights\"]) )\n",
    "            #print(\"VVT' :\", VVT.get_shape(), file = sys.stderr)\n",
    "            x_M = tf.matmul(self.vars.x, VVT)  \n",
    "            \n",
    "            self.vars[\"term2\"] =  tf.reshape(tf.batch_matmul(\n",
    "                    tf.reshape(x_M, [-1,  1, self.xlen,] ), \n",
    "                    tf.reshape(self.vars.x, [-1, self.xlen, 1 ] ),\n",
    "                    ), [-1,1])\n",
    "            #print(\"term2 :\", self.vars[\"term2\"].get_shape(), file = sys.stderr)\n",
    "            self.vars[\"term1\"]  = tf.matmul( self.vars.x, tf.transpose(self.parameters[\"weights\"]))\n",
    "            self.vars.y_predicted = tf.nn.sigmoid(  self.parameters[\"bias\"] + self.vars.term1 + self.vars.term2 )\n",
    "            \n",
    "            \n",
    "        self.saver = tf.train.Saver()\n",
    "        return\n",
    "    \n",
    "#     def _create_loss(self):\n",
    "#         # Minimize the squared errors\n",
    "#         l2_loss = tf.reduce_mean(tf.pow( self.vars.y_predicted - self.vars.yy, 2))\n",
    "#         l2_sy = tf.scalar_summary( \"L2_loss\", l2_loss )\n",
    "#         \"Lasso penalty\"\n",
    "#         l1_penalty = tf.reduce_sum((tf.abs(tf.concat(1, [self.W1,self.b1], name = \"l1\" ) )) )\n",
    "#         l1p_sy =  tf.scalar_summary( \"L1_penalty\" , l1_penalty )\n",
    "#         \"total\"\n",
    "#         tot_loss = l2_loss + self.ALPHA * l1_penalty\n",
    "#         tot_loss_sy =  tf.scalar_summary( \"loss\" , tot_loss )\n",
    "#         \"R2\"\n",
    "#         _, y_var = tf.nn.moments(self.vars.yy, [0,1])\n",
    "#         rsq =  1 - l2_loss / y_var\n",
    "#         rsq_sy = tf.scalar_summary( \"R2\", rsq)\n",
    "#         return tot_loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 253,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(4, 9)\n"
     ]
    }
   ],
   "source": [
    "from sklearn.feature_extraction import DictVectorizer\n",
    "train_X = [\n",
    "    {\"user\": \"1\", \"item\": \"5\", \"age\": 19},\n",
    "    {\"user\": \"2\", \"item\": \"43\", \"age\": 33},\n",
    "    {\"user\": \"3\", \"item\": \"20\", \"age\": 55},\n",
    "    {\"user\": \"4\", \"item\": \"10\", \"age\": 20},\n",
    "]\n",
    "v = DictVectorizer()\n",
    "train_X = v.fit_transform(train_X)\n",
    "print( train_X.toarray().shape )\n",
    "\n",
    "train_Y = np.repeat(1.0, train_X.shape[0])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## fit"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 254,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "variance(y) =  0.0\n",
      "loading a session\n",
      "training epochs: 18000 ... 28000, saving each 1000' epoch\n",
      "Epoch: 18000\t\n",
      "\ttrain\tloss: 0.0000\tL2_loss: 0.0000\tR2: -inf\n",
      "Epoch: 19000\t\n",
      "\ttrain\tloss: 0.0000\tL2_loss: 0.0000\tR2: -inf\n",
      "Epoch: 20000\t\n",
      "\ttrain\tloss: 0.0000\tL2_loss: 0.0000\tR2: -inf\n",
      "Epoch: 21000\t\n",
      "\ttrain\tloss: 0.0000\tL2_loss: 0.0000\tR2: -inf\n",
      "Epoch: 22000\t\n",
      "\ttrain\tloss: 0.0000\tL2_loss: 0.0000\tR2: -inf\n",
      "Epoch: 23000\t\n",
      "\ttrain\tloss: 0.0000\tL2_loss: 0.0000\tR2: -inf\n",
      "Epoch: 24000\t\n",
      "\ttrain\tloss: 0.0000\tL2_loss: 0.0000\tR2: -inf\n",
      "Epoch: 25000\t\n",
      "\ttrain\tloss: 0.0000\tL2_loss: 0.0000\tR2: -inf\n",
      "Epoch: 26000\t\n",
      "\ttrain\tloss: 0.0000\tL2_loss: 0.0000\tR2: -inf\n",
      "Epoch: 27000\t\n",
      "\ttrain\tloss: 0.0000\tL2_loss: 0.0000\tR2: -inf\n",
      "Optimization Finished!\n",
      "_ipython_display_ not found\n",
      "_repr_javascript_ not found\n",
      "_repr_pdf_ not found\n",
      "_repr_html_ not found\n",
      "_repr_json_ not found\n",
      "_repr_png_ not found\n",
      "_repr_svg_ not found\n",
      "_repr_jpeg_ not found\n",
      "_repr_markdown_ not found\n",
      "_repr_latex_ not found\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<__main__.tffactmachine at 0x7fb7150dc278>"
      ]
     },
     "execution_count": 254,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tffm = tffactmachine( display_step = 1000, BATCH_SIZE = 40, dropout = 0,\n",
    "            # optimizer = tf.train.GradientDescentOptimizer,\n",
    "             learning_rate = 1e-4, checkpoint_dir = \"./factmachine/\", num_factors=10)\n",
    "tffm.fit(train_X.toarray(), train_Y, load = True, training_epochs = 10000,)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 255,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "loading a session\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([[ 0.99981469],\n",
       "       [ 0.99999988],\n",
       "       [ 0.91249973],\n",
       "       [ 1.        ]], dtype=float32)"
      ]
     },
     "execution_count": 255,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_X =v.transform([{\"user\": \"1\", \"item\": \"10\", \"age\": 24},\n",
    "                    {\"user\": \"2\", \"item\": \"43\", \"age\": 33},\n",
    "                      {\"user\": \"1\", \"item\": \"10\", \"age\": 12},\n",
    "                     {\"user\": \"1\", \"item\": \"20\", \"age\": 59}]).toarray()\n",
    "#print(test_X)\n",
    "ypredicted = tffm.predict(test_X, debug = False)\n",
    "#print(\"shape: \", ypredicted.shape)\n",
    "ypredicted"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 256,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.text.Text at 0x7fb719d3e630>"
      ]
     },
     "execution_count": 256,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZoAAAEZCAYAAACuIuMVAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fb719d322b0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "t_ = [x*20 for x in range(len(tffm.train_summary))]\n",
    "plt.plot( t_, [x[\"loss\"] for x in tffm.train_summary ] , \"r\")\n",
    "plt.ylim([0, 0.001* np.ceil(10*max([x[\"loss\"] for x in tffm.train_summary]))])\n",
    "plt.ylabel(\"loss\")\n",
    "plt.xlabel(\"epoch\")\n",
    "plt.title(\"learning curve\")\n",
    "#plt.savefig(\"progress.eps\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## visualize weights"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 257,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "loading a session\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "dict_keys(['weights', 'bias', 'veights'])"
      ]
     },
     "execution_count": 257,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "params = tffm.get_params(load = True)\n",
    "params.keys()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 258,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYAAAAEACAYAAAC6d6FnAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fb717046ac8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(params[\"veights\"].T, \"g.-\", color = (0,1,0,0.8))\n",
    "pass"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.4.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}