Softmax.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Example code for Softmax\n",
    "=========================="
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0 1.09774 [[ -8.33333252e-05   4.16666626e-05   4.16666480e-05]\n",
      " [  1.66666694e-04   2.91666773e-04  -4.58333408e-04]\n",
      " [  1.66666636e-04   4.16666706e-04  -5.83333429e-04]]\n",
      "100 1.06645 [[-0.00974582  0.00085278  0.00889304]\n",
      " [ 0.00999095  0.01240054 -0.02239149]\n",
      " [ 0.01006747  0.02499313 -0.0350606 ]]\n",
      "200 1.05962 [[-0.02051384 -0.00103983  0.02155367]\n",
      " [ 0.01406438  0.01097753 -0.02504191]\n",
      " [ 0.01431208  0.03574873 -0.05006079]]\n",
      "300 1.0546 [[-0.03162513 -0.00356464  0.03518977]\n",
      " [ 0.01614517  0.00659079 -0.02273595]\n",
      " [ 0.01664983  0.04315857 -0.05980836]]\n",
      "400 1.04985 [[-0.04282598 -0.00625899  0.04908497]\n",
      " [ 0.01747187  0.00156368 -0.01903554]\n",
      " [ 0.01831204  0.04954104 -0.06785304]]\n",
      "500 1.04523 [[-0.05402156 -0.00900351  0.06302506]\n",
      " [ 0.01851289 -0.00351735 -0.01499552]\n",
      " [ 0.01976016  0.05549131 -0.07525142]]\n",
      "600 1.0407 [[-0.06517859 -0.01176361  0.07694222]\n",
      " [ 0.01943521 -0.00848972 -0.01094548]\n",
      " [ 0.0211558   0.06118308 -0.0823388 ]]\n",
      "700 1.03628 [[-0.07628597 -0.01452834  0.09081434]\n",
      " [ 0.02029658 -0.01330799 -0.00698858]\n",
      " [ 0.02255237  0.0666732  -0.08922548]]\n",
      "800 1.03194 [[-0.08734013 -0.01729389  0.10463405]\n",
      " [ 0.0211172  -0.01796171 -0.00315548]\n",
      " [ 0.02396628  0.07198346 -0.09594974]]\n",
      "900 1.0277 [[-0.09834011 -0.02005867  0.11839879]\n",
      " [ 0.02190469 -0.02245126  0.00054658]\n",
      " [ 0.02540163  0.07712454 -0.10252618]]\n",
      "1000 1.02354 [[-0.10928574 -0.02282181  0.13210757]\n",
      " [ 0.02266254 -0.02678034  0.00411783]\n",
      " [ 0.02685853  0.08210357 -0.10896212]]\n",
      "1100 1.01947 [[-0.12017717 -0.02558274  0.14575994]\n",
      " [ 0.02339279 -0.03095369  0.00756093]\n",
      " [ 0.02833582  0.08692633 -0.11526213]]\n",
      "1200 1.01547 [[-0.13101467 -0.02834092  0.15935561]\n",
      " [ 0.02409703 -0.03497621  0.01087923]\n",
      " [ 0.029832    0.091598   -0.12143001]]\n",
      "1300 1.01155 [[-0.14179853 -0.03109588  0.1728944 ]\n",
      " [ 0.02477667 -0.03885284  0.01407623]\n",
      " [ 0.03134547  0.09612366 -0.12746911]]\n",
      "1400 1.0077 [[-0.15252906 -0.03384715  0.18637618]\n",
      " [ 0.02543302 -0.04258838  0.01715542]\n",
      " [ 0.03287466  0.10050805 -0.13338271]]\n",
      "1500 1.00392 [[-0.16320656 -0.03659431  0.19980085]\n",
      " [ 0.02606733 -0.04618748  0.02012023]\n",
      " [ 0.03441809  0.10475579 -0.13917387]]\n",
      "1600 1.00021 [[-0.17383142 -0.03933692  0.21316831]\n",
      " [ 0.02668081 -0.04965464  0.0229739 ]\n",
      " [ 0.03597427  0.10887136 -0.14484565]]\n",
      "1700 0.996564 [[-0.18440396 -0.04207458  0.22647853]\n",
      " [ 0.02727464 -0.05299426  0.0257197 ]\n",
      " [ 0.03754187  0.11285908 -0.15040097]]\n",
      "1800 0.992979 [[-0.19492456 -0.04480688  0.23973146]\n",
      " [ 0.02784993 -0.05621058  0.02836074]\n",
      " [ 0.03911954  0.11672314 -0.15584265]]\n",
      "1900 0.989455 [[-0.2053936  -0.04753346  0.25292706]\n",
      " [ 0.02840775 -0.0593077   0.03090005]\n",
      " [ 0.04070605  0.12046755 -0.16117358]]\n",
      "2000 0.985988 [[-0.21581143 -0.05025396  0.26606542]\n",
      " [ 0.02894915 -0.06228962  0.03334056]\n",
      " [ 0.04230019  0.12409624 -0.16639642]]\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "\n",
    "# Load data\n",
    "xy = np.loadtxt('softmax_train.txt', unpack=True, dtype='float32')\n",
    "x_data = np.transpose(xy[0:3])\n",
    "y_data = np.transpose(xy[3:])\n",
    "\n",
    "# Graph input\n",
    "X = tf.placeholder(\"float\", [None,3] )\n",
    "Y = tf.placeholder(\"float\", [None,3] )\n",
    "\n",
    "# Model weight\n",
    "W = tf.Variable(tf.zeros([3,3]))\n",
    "\n",
    "# Construct model\n",
    "hypothesis = tf.nn.softmax( tf.matmul(X,W))\n",
    "\n",
    "# Learning rate\n",
    "learning_rate = 0.001\n",
    "\n",
    "# Cross entrophy\n",
    "cost = tf.reduce_mean( -tf.reduce_sum( Y*tf.log(hypothesis), reduction_indices=1 ) )\n",
    "\n",
    "# Gradient Descent\n",
    "optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(cost)\n",
    "\n",
    "# Initializing the variables\n",
    "init = tf.initialize_all_variables()\n",
    "\n",
    "# Launch graph\n",
    "sess = tf.Session()\n",
    "sess.run(init)\n",
    "\n",
    "cost_list = []\n",
    "step_list = []\n",
    "for step in range(2001):        \n",
    "    sess.run(optimizer, feed_dict={X:x_data, Y:y_data})\n",
    "    if step % 100 == 0:\n",
    "        cost_in_step = sess.run(cost, feed_dict={X:x_data, Y:y_data})\n",
    "        print(step, cost_in_step, sess.run(W) )\n",
    "        step_list.append(step)\n",
    "        cost_list.append(cost_in_step)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Evaluation of model\n",
    "====================="
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[ 0.46272627  0.35483006  0.18244371]] [0]\n",
      "[[ 0.33820099  0.42101386  0.24078514]] [1]\n",
      "[[ 0.27002314  0.29085544  0.4391214 ]] [2]\n"
     ]
    }
   ],
   "source": [
    "a = sess.run(hypothesis, feed_dict={X:[[1, 11, 7]]} )\n",
    "print(a, sess.run(tf.arg_max(a,1)) )\n",
    "\n",
    "b = sess.run(hypothesis, feed_dict={X:[[1, 3, 4]]} )\n",
    "print(b, sess.run(tf.arg_max(b,1)) )\n",
    "\n",
    "c = sess.run(hypothesis, feed_dict={X:[[1, 1, 0]]} )\n",
    "print(c, sess.run(tf.arg_max(c,1)) )\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Chart the cost reduction in every step.\n",
    "==============\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYQAAAEACAYAAACznAEdAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x114ed52b0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "\n",
    "import matplotlib\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.plot(step_list, cost_list, label='Cost function in every step.')\n",
    "plt.legend()\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[-0.21581143 -0.05025396  0.26606542]\n",
      " [ 0.02894915 -0.06228962  0.03334056]\n",
      " [ 0.04230019  0.12409624 -0.16639642]]\n"
     ]
    }
   ],
   "source": [
    "# Tip: print out the tensor.\n",
    "print( sess.run(W) )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.5.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}