{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import numpy.random as rng\n",
    "import tensorflow as tf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "n = 1000\n",
    "din = np.random.randint(0,2, (n, 2)).astype(bool)\n",
    "out = np.logical_xor(din[:,0], din[:,1])\n",
    "\n",
    "out = np.zeros((n, 2), dtype=\"bool\")\n",
    "\n",
    "out[:,1] = np.logical_xor(din[:,0], din[:,1])\n",
    "out[:,0] = True ^ out[:,1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "x = tf.placeholder(\"bool\", shape=[None,2])\n",
    "x1 = tf.cast(x, \"float\")\n",
    "\n",
    "with tf.name_scope('Layer1'):\n",
    "    W1 = tf.Variable(tf.random_normal([2,10]))\n",
    "    b1 = tf.Variable(tf.random_normal([10]))\n",
    "    hidden = tf.nn.relu(tf.matmul(x1, W1) + b1)\n",
    "    \n",
    "    \n",
    "with tf.name_scope('Layer2'):\n",
    "    W2 = tf.Variable(tf.random_normal([10,2]))\n",
    "    b2 = tf.Variable(tf.random_normal([2]))\n",
    "    hidden1 =  tf.matmul(hidden, W2) + b2\n",
    "\n",
    "with tf.name_scope('output'):\n",
    "    y = tf.nn.softmax(hidden1)\n",
    "    y_val = tf.arg_max(y,1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "y_ = tf.placeholder(\"float\",[None,2])\n",
    "\n",
    "loss = tf.reduce_sum(-y_*tf.log(y))\n",
    "init = tf.initialize_all_variables()\n",
    "optimizer = tf.train.GradientDescentOptimizer(.001).minimize(loss)\n",
    "\n",
    "\n",
    "tf.scalar_summary(\"cost\", loss)\n",
    "\n",
    "merged = tf.merge_all_summaries()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3329.73\n",
      "Iteration 0 Loss 3329.73\n",
      "Iteration 10 Loss 277.95\n",
      "Iteration 20 Loss 126.06\n",
      "Iteration 30 Loss 72.0056\n"
     ]
    }
   ],
   "source": [
    "fd = {x: din, y_:out}\n",
    "with tf.Session() as sess:\n",
    "    sess.run(init)\n",
    "    summary_writer = tf.train.SummaryWriter(\"./logs/run1\")\n",
    "    summary_writer.add_graph(y.graph)\n",
    "    \n",
    "    print(sess.run(loss, feed_dict=fd))\n",
    "    \n",
    "    \n",
    "    \n",
    "    for step in range(40):\n",
    "        summ, cost_curr, _ = sess.run([merged, loss, optimizer], feed_dict={x: din, y_: out})\n",
    "        if step % 10 == 0 : print(\"Iteration\", step, \"Loss\", cost_curr)\n",
    "            \n",
    "        summary_writer.add_summary(summ, step)\n",
    "            \n",
    "            \n",
    "#     print(sess.run(y_val, feed_dict={x:[[0,0], [1,1], [1,0], [0,1]]}))"
   ]
  },
  {
   "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
}