Merge pull request #1 from energyfirefox/energyfirefox-patch-1

Fixed ValueError during model creation.

Author: energyfirefox

Concept02_rnn.ipynb

@@ -0,0 +1,202 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# Ch `10`: Concept `02`"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Recurrent Neural Network"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Import the relevant libraries:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "/Users/anastasiia/anaconda3/lib/python3.6/site-packages/h5py/__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.\n",
+      "  from ._conv import register_converters as _register_converters\n"
+     ]
+    }
+   ],
+   "source": [
+    "import numpy as np\n",
+    "import tensorflow as tf\n",
+    "from tensorflow.contrib import rnn"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Define the RNN model:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "class SeriesPredictor:\n",
+    "\n",
+    "    def __init__(self, input_dim, seq_size, hidden_dim=10):\n",
+    "        # Hyperparameters\n",
+    "        self.input_dim = input_dim\n",
+    "        self.seq_size = seq_size\n",
+    "        self.hidden_dim = hidden_dim\n",
+    "\n",
+    "        # Weight variables and input placeholders\n",
+    "        self.W_out = tf.Variable(tf.random_normal([hidden_dim, 1]), name='W_out')\n",
+    "        self.b_out = tf.Variable(tf.random_normal([1]), name='b_out')\n",
+    "        self.x = tf.placeholder(tf.float32, [None, seq_size, input_dim])\n",
+    "        self.y = tf.placeholder(tf.float32, [None, seq_size])\n",
+    "\n",
+    "        # Cost optimizer\n",
+    "        self.cost = tf.reduce_mean(tf.square(self.model() - self.y))\n",
+    "        self.train_op = tf.train.AdamOptimizer().minimize(self.cost)\n",
+    "\n",
+    "        # Auxiliary ops\n",
+    "        self.saver = tf.train.Saver()\n",
+    "\n",
+    "    def model(self):\n",
+    "        \"\"\"\n",
+    "        :param x: inputs of size [T, batch_size, input_size]\n",
+    "        :param W: matrix of fully-connected output layer weights\n",
+    "        :param b: vector of fully-connected output layer biases\n",
+    "        \"\"\"\n",
+    "        cell = rnn.BasicLSTMCell(self.hidden_dim, reuse=tf.get_variable_scope().reuse)\n",
+    "        outputs, states = tf.nn.dynamic_rnn(cell, self.x, dtype=tf.float32)\n",
+    "        num_examples = tf.shape(self.x)[0]\n",
+    "        W_repeated = tf.tile(tf.expand_dims(self.W_out, 0), [num_examples, 1, 1])\n",
+    "        out = tf.matmul(outputs, W_repeated) + self.b_out\n",
+    "        out = tf.squeeze(out)\n",
+    "        return out\n",
+    "\n",
+    "    def train(self, train_x, train_y):\n",
+    "        with tf.Session() as sess:\n",
+    "            tf.get_variable_scope().reuse_variables()\n",
+    "            sess.run(tf.global_variables_initializer())\n",
+    "            for i in range(1000):\n",
+    "                _, mse = sess.run([self.train_op, self.cost], feed_dict={self.x: train_x, self.y: train_y})\n",
+    "                if i % 100 == 0:\n",
+    "                    print(i, mse)\n",
+    "            save_path = self.saver.save(sess, 'model.ckpt')\n",
+    "            print('Model saved to {}'.format(save_path))\n",
+    "\n",
+    "    def test(self, test_x):\n",
+    "        with tf.Session() as sess:\n",
+    "            tf.get_variable_scope().reuse_variables()\n",
+    "            self.saver.restore(sess, './model.ckpt')\n",
+    "            output = sess.run(self.model(), feed_dict={self.x: test_x})\n",
+    "            return output\n",
+    "\n",
+    "\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Now, we'll train a series predictor. Let's say we have a sequence of numbers `[a, b, c, d]` that we want to transform into `[a, a+b, b+c, c+d]`. We'll give the RNN a couple examples in the training data. Let's see how well it learns this intended transformation:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "0 70.91218\n",
+      "100 28.614563\n",
+      "200 10.268018\n",
+      "300 5.664831\n",
+      "400 3.5102208\n",
+      "500 1.883831\n",
+      "600 1.0723968\n",
+      "700 0.6708067\n",
+      "800 0.4667667\n",
+      "900 0.35269937\n",
+      "Model saved to model.ckpt\n",
+      "INFO:tensorflow:Restoring parameters from ./model.ckpt\n",
+      "\n",
+      "Lets run some tests!\n",
+      "\n",
+      "When the input is [[1], [2], [3], [4]]\n",
+      "The ground truth output should be [[1], [3], [5], [7]]\n",
+      "And the model thinks it is [1.6758004 2.7610283 4.739178  7.087058 ]\n",
+      "\n",
+      "When the input is [[4], [5], [6], [7]]\n",
+      "The ground truth output should be [[4], [9], [11], [13]]\n",
+      "And the model thinks it is [ 4.4391885  9.112013  12.074081  13.157787 ]\n",
+      "\n"
+     ]
+    }
+   ],
+   "source": [
+    "if __name__ == '__main__':\n",
+    "    predictor = SeriesPredictor(input_dim=1, seq_size=4, hidden_dim=10)\n",
+    "    train_x = [[[1], [2], [5], [6]],\n",
+    "               [[5], [7], [7], [8]],\n",
+    "               [[3], [4], [5], [7]]]\n",
+    "    train_y = [[1, 3, 7, 11],\n",
+    "               [5, 12, 14, 15],\n",
+    "               [3, 7, 9, 12]]\n",
+    "    predictor.train(train_x, train_y)\n",
+    "\n",
+    "    test_x = [[[1], [2], [3], [4]],  # 1, 3, 5, 7\n",
+    "              [[4], [5], [6], [7]]]  # 4, 9, 11, 13\n",
+    "    actual_y = [[[1], [3], [5], [7]],\n",
+    "                [[4], [9], [11], [13]]]\n",
+    "    pred_y = predictor.test(test_x)\n",
+    "    \n",
+    "    print(\"\\nLets run some tests!\\n\")\n",
+    "    \n",
+    "    for i, x in enumerate(test_x):\n",
+    "        print(\"When the input is {}\".format(x))\n",
+    "        print(\"The ground truth output should be {}\".format(actual_y[i]))\n",
+    "        print(\"And the model thinks it is {}\\n\".format(pred_y[i]))"
+   ]
+  }
+ ],
+ "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": "3.6.5"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 1
+}