From 6a98b3235f916d6e3845943080699924db090774 Mon Sep 17 00:00:00 2001 From: Donne Martin Date: Mon, 28 Dec 2015 07:54:42 -0500 Subject: [PATCH] Add TensorFlow AlexNet notebook. --- README.md | 1 + .../notebooks/3_neural_networks/alexnet.ipynb | 345 ++++++++++++++++++ 2 files changed, 346 insertions(+) create mode 100644 deep-learning/tensor-flow-examples/notebooks/3_neural_networks/alexnet.ipynb diff --git a/README.md b/README.md index da54e5bc..1e9e0d78 100644 --- a/README.md +++ b/README.md @@ -98,6 +98,7 @@ IPython Notebook(s) demonstrating deep learning functionality. | [tsf-linear](http://nbviewer.ipython.org/github/donnemartin/data-science-ipython-notebooks/blob/master/deep-learning/tensor-flow-examples/2_basic_classifiers/linear_regression.ipynb) | Implement linear regression in TensorFlow. | | [tsf-logistic](http://nbviewer.ipython.org/github/donnemartin/data-science-ipython-notebooks/blob/master/deep-learning/tensor-flow-examples/2_basic_classifiers/logistic_regression.ipynb) | Implement logistic regression in TensorFlow. | | [tsf-nn](http://nbviewer.ipython.org/github/donnemartin/data-science-ipython-notebooks/blob/master/deep-learning/tensor-flow-examples/2_basic_classifiers/nearest_neighbor.ipynb) | Implement nearest neighboars in TensorFlow. | +| [tsf-alex](http://nbviewer.ipython.org/github/donnemartin/data-science-ipython-notebooks/blob/master/deep-learning/tensor-flow-examples/3_neural_networks/alexnet.ipynb) | Implement AlexNet in TensorFlow. | ### tensor-flow-exercises diff --git a/deep-learning/tensor-flow-examples/notebooks/3_neural_networks/alexnet.ipynb b/deep-learning/tensor-flow-examples/notebooks/3_neural_networks/alexnet.ipynb new file mode 100644 index 00000000..d3b600de --- /dev/null +++ b/deep-learning/tensor-flow-examples/notebooks/3_neural_networks/alexnet.ipynb @@ -0,0 +1,345 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# AlexNet in TensorFlow\n", + "\n", + "Credits: Forked from [TensorFlow-Examples](https://github.com/aymericdamien/TensorFlow-Examples) by Aymeric Damien\n", + "\n", + "## Setup\n", + "\n", + "Refer to the [setup instructions](http://nbviewer.ipython.org/github/donnemartin/data-science-ipython-notebooks/blob/master/deep-learning/tensor-flow-examples/Setup_TensorFlow.md)" + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "metadata": { + "collapsed": false + }, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Extracting /tmp/data/train-images-idx3-ubyte.gz\n", + "Extracting /tmp/data/train-labels-idx1-ubyte.gz\n", + "Extracting /tmp/data/t10k-images-idx3-ubyte.gz\n", + "Extracting /tmp/data/t10k-labels-idx1-ubyte.gz\n" + ] + } + ], + "source": [ + "# Import MINST data\n", + "import input_data\n", + "mnist = input_data.read_data_sets(\"/tmp/data/\", one_hot=True)" + ] + }, + { + "cell_type": "code", + "execution_count": 3, + "metadata": { + "collapsed": true + }, + "outputs": [], + "source": [ + "import tensorflow as tf" + ] + }, + { + "cell_type": "code", + "execution_count": 17, + "metadata": { + "collapsed": true + }, + "outputs": [], + "source": [ + "# Parameters\n", + "learning_rate = 0.001\n", + "training_iters = 300000\n", + "batch_size = 64\n", + "display_step = 100" + ] + }, + { + "cell_type": "code", + "execution_count": 5, + "metadata": { + "collapsed": true + }, + "outputs": [], + "source": [ + "# Network Parameters\n", + "n_input = 784 # MNIST data input (img shape: 28*28)\n", + "n_classes = 10 # MNIST total classes (0-9 digits)\n", + "dropout = 0.8 # Dropout, probability to keep units" + ] + }, + { + "cell_type": "code", + "execution_count": 6, + "metadata": { + "collapsed": true + }, + "outputs": [], + "source": [ + "# tf Graph input\n", + "x = tf.placeholder(tf.types.float32, [None, n_input])\n", + "y = tf.placeholder(tf.types.float32, [None, n_classes])\n", + "keep_prob = tf.placeholder(tf.types.float32) # dropout (keep probability)" + ] + }, + { + "cell_type": "code", + "execution_count": 7, + "metadata": { + "collapsed": true + }, + "outputs": [], + "source": [ + "# Create AlexNet model\n", + "def conv2d(name, l_input, w, b):\n", + " return tf.nn.relu(tf.nn.bias_add(tf.nn.conv2d(l_input, w, strides=[1, 1, 1, 1], \n", + " padding='SAME'),b), name=name)\n", + "\n", + "def max_pool(name, l_input, k):\n", + " return tf.nn.max_pool(l_input, ksize=[1, k, k, 1], strides=[1, k, k, 1], \n", + " padding='SAME', name=name)\n", + "\n", + "def norm(name, l_input, lsize=4):\n", + " return tf.nn.lrn(l_input, lsize, bias=1.0, alpha=0.001 / 9.0, beta=0.75, name=name)\n", + "\n", + "def alex_net(_X, _weights, _biases, _dropout):\n", + " # Reshape input picture\n", + " _X = tf.reshape(_X, shape=[-1, 28, 28, 1])\n", + "\n", + " # Convolution Layer\n", + " conv1 = conv2d('conv1', _X, _weights['wc1'], _biases['bc1'])\n", + " # Max Pooling (down-sampling)\n", + " pool1 = max_pool('pool1', conv1, k=2)\n", + " # Apply Normalization\n", + " norm1 = norm('norm1', pool1, lsize=4)\n", + " # Apply Dropout\n", + " norm1 = tf.nn.dropout(norm1, _dropout)\n", + "\n", + " # Convolution Layer\n", + " conv2 = conv2d('conv2', norm1, _weights['wc2'], _biases['bc2'])\n", + " # Max Pooling (down-sampling)\n", + " pool2 = max_pool('pool2', conv2, k=2)\n", + " # Apply Normalization\n", + " norm2 = norm('norm2', pool2, lsize=4)\n", + " # Apply Dropout\n", + " norm2 = tf.nn.dropout(norm2, _dropout)\n", + "\n", + " # Convolution Layer\n", + " conv3 = conv2d('conv3', norm2, _weights['wc3'], _biases['bc3'])\n", + " # Max Pooling (down-sampling)\n", + " pool3 = max_pool('pool3', conv3, k=2)\n", + " # Apply Normalization\n", + " norm3 = norm('norm3', pool3, lsize=4)\n", + " # Apply Dropout\n", + " norm3 = tf.nn.dropout(norm3, _dropout)\n", + "\n", + " # Fully connected layer\n", + " # Reshape conv3 output to fit dense layer input\n", + " dense1 = tf.reshape(norm3, [-1, _weights['wd1'].get_shape().as_list()[0]]) \n", + " # Relu activation\n", + " dense1 = tf.nn.relu(tf.matmul(dense1, _weights['wd1']) + _biases['bd1'], name='fc1')\n", + " \n", + " # Relu activation\n", + " dense2 = tf.nn.relu(tf.matmul(dense1, _weights['wd2']) + _biases['bd2'], name='fc2') \n", + "\n", + " # Output, class prediction\n", + " out = tf.matmul(dense2, _weights['out']) + _biases['out']\n", + " return out" + ] + }, + { + "cell_type": "code", + "execution_count": 8, + "metadata": { + "collapsed": true + }, + "outputs": [], + "source": [ + "# Store layers weight & bias\n", + "weights = {\n", + " 'wc1': tf.Variable(tf.random_normal([3, 3, 1, 64])),\n", + " 'wc2': tf.Variable(tf.random_normal([3, 3, 64, 128])),\n", + " 'wc3': tf.Variable(tf.random_normal([3, 3, 128, 256])),\n", + " 'wd1': tf.Variable(tf.random_normal([4*4*256, 1024])),\n", + " 'wd2': tf.Variable(tf.random_normal([1024, 1024])),\n", + " 'out': tf.Variable(tf.random_normal([1024, 10]))\n", + "}\n", + "biases = {\n", + " 'bc1': tf.Variable(tf.random_normal([64])),\n", + " 'bc2': tf.Variable(tf.random_normal([128])),\n", + " 'bc3': tf.Variable(tf.random_normal([256])),\n", + " 'bd1': tf.Variable(tf.random_normal([1024])),\n", + " 'bd2': tf.Variable(tf.random_normal([1024])),\n", + " 'out': tf.Variable(tf.random_normal([n_classes]))\n", + "}" + ] + }, + { + "cell_type": "code", + "execution_count": 9, + "metadata": { + "collapsed": true + }, + "outputs": [], + "source": [ + "# Construct model\n", + "pred = alex_net(x, weights, biases, keep_prob)" + ] + }, + { + "cell_type": "code", + "execution_count": 10, + "metadata": { + "collapsed": true + }, + "outputs": [], + "source": [ + "# Define loss and optimizer\n", + "cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(pred, y))\n", + "optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate).minimize(cost)" + ] + }, + { + "cell_type": "code", + "execution_count": 11, + "metadata": { + "collapsed": true + }, + "outputs": [], + "source": [ + "# Evaluate model\n", + "correct_pred = tf.equal(tf.argmax(pred,1), tf.argmax(y,1))\n", + "accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.types.float32))" + ] + }, + { + "cell_type": "code", + "execution_count": 12, + "metadata": { + "collapsed": true + }, + "outputs": [], + "source": [ + "# Initializing the variables\n", + "init = tf.initialize_all_variables()" + ] + }, + { + "cell_type": "code", + "execution_count": 18, + "metadata": { + "collapsed": false + }, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Iter 6400, Minibatch Loss= 29666.185547, Training Accuracy= 0.59375\n", + "Iter 12800, Minibatch Loss= 22125.562500, Training Accuracy= 0.60938\n", + "Iter 19200, Minibatch Loss= 22631.134766, Training Accuracy= 0.59375\n", + "Iter 25600, Minibatch Loss= 18498.414062, Training Accuracy= 0.62500\n", + "Iter 32000, Minibatch Loss= 11318.283203, Training Accuracy= 0.70312\n", + "Iter 38400, Minibatch Loss= 12076.280273, Training Accuracy= 0.70312\n", + "Iter 44800, Minibatch Loss= 8195.520508, Training Accuracy= 0.82812\n", + "Iter 51200, Minibatch Loss= 5176.181641, Training Accuracy= 0.84375\n", + "Iter 57600, Minibatch Loss= 8951.896484, Training Accuracy= 0.81250\n", + "Iter 64000, Minibatch Loss= 10096.946289, Training Accuracy= 0.78125\n", + "Iter 70400, Minibatch Loss= 11466.641602, Training Accuracy= 0.68750\n", + "Iter 76800, Minibatch Loss= 7469.824219, Training Accuracy= 0.78125\n", + "Iter 83200, Minibatch Loss= 4147.449219, Training Accuracy= 0.89062\n", + "Iter 89600, Minibatch Loss= 5904.782227, Training Accuracy= 0.82812\n", + "Iter 96000, Minibatch Loss= 718.493713, Training Accuracy= 0.93750\n", + "Iter 102400, Minibatch Loss= 2184.151367, Training Accuracy= 0.93750\n", + "Iter 108800, Minibatch Loss= 2354.463135, Training Accuracy= 0.89062\n", + "Iter 115200, Minibatch Loss= 8612.959961, Training Accuracy= 0.81250\n", + "Iter 121600, Minibatch Loss= 2225.773926, Training Accuracy= 0.84375\n", + "Iter 128000, Minibatch Loss= 160.583618, Training Accuracy= 0.96875\n", + "Iter 134400, Minibatch Loss= 1524.846069, Training Accuracy= 0.93750\n", + "Iter 140800, Minibatch Loss= 3501.871094, Training Accuracy= 0.89062\n", + "Iter 147200, Minibatch Loss= 661.977051, Training Accuracy= 0.96875\n", + "Iter 153600, Minibatch Loss= 367.857788, Training Accuracy= 0.98438\n", + "Iter 160000, Minibatch Loss= 1735.458740, Training Accuracy= 0.90625\n", + "Iter 166400, Minibatch Loss= 209.320374, Training Accuracy= 0.95312\n", + "Iter 172800, Minibatch Loss= 1788.553955, Training Accuracy= 0.90625\n", + "Iter 179200, Minibatch Loss= 912.995544, Training Accuracy= 0.93750\n", + "Iter 185600, Minibatch Loss= 2534.074463, Training Accuracy= 0.87500\n", + "Iter 192000, Minibatch Loss= 73.052612, Training Accuracy= 0.96875\n", + "Iter 198400, Minibatch Loss= 1609.606323, Training Accuracy= 0.93750\n", + "Iter 204800, Minibatch Loss= 1823.219727, Training Accuracy= 0.96875\n", + "Iter 211200, Minibatch Loss= 578.051086, Training Accuracy= 0.96875\n", + "Iter 217600, Minibatch Loss= 1532.326172, Training Accuracy= 0.89062\n", + "Iter 224000, Minibatch Loss= 769.775269, Training Accuracy= 0.95312\n", + "Iter 230400, Minibatch Loss= 2614.737793, Training Accuracy= 0.92188\n", + "Iter 236800, Minibatch Loss= 938.664368, Training Accuracy= 0.95312\n", + "Iter 243200, Minibatch Loss= 1520.495605, Training Accuracy= 0.93750\n", + "Iter 249600, Minibatch Loss= 657.419739, Training Accuracy= 0.95312\n", + "Iter 256000, Minibatch Loss= 522.802124, Training Accuracy= 0.90625\n", + "Iter 262400, Minibatch Loss= 211.188477, Training Accuracy= 0.96875\n", + "Iter 268800, Minibatch Loss= 520.451172, Training Accuracy= 0.92188\n", + "Iter 275200, Minibatch Loss= 1418.759155, Training Accuracy= 0.89062\n", + "Iter 281600, Minibatch Loss= 241.748596, Training Accuracy= 0.96875\n", + "Iter 288000, Minibatch Loss= 0.000000, Training Accuracy= 1.00000\n", + "Iter 294400, Minibatch Loss= 1535.772827, Training Accuracy= 0.92188\n", + "Optimization Finished!\n", + "Testing Accuracy: 0.980469\n" + ] + } + ], + "source": [ + "# Launch the graph\n", + "with tf.Session() as sess:\n", + " sess.run(init)\n", + " step = 1\n", + " # Keep training until reach max iterations\n", + " while step * batch_size < training_iters:\n", + " batch_xs, batch_ys = mnist.train.next_batch(batch_size)\n", + " # Fit training using batch data\n", + " sess.run(optimizer, feed_dict={x: batch_xs, y: batch_ys, keep_prob: dropout})\n", + " if step % display_step == 0:\n", + " # Calculate batch accuracy\n", + " acc = sess.run(accuracy, feed_dict={x: batch_xs, y: batch_ys, keep_prob: 1.})\n", + " # Calculate batch loss\n", + " loss = sess.run(cost, feed_dict={x: batch_xs, y: batch_ys, keep_prob: 1.})\n", + " print \"Iter \" + str(step*batch_size) + \", Minibatch Loss= \" \\\n", + " + \"{:.6f}\".format(loss) + \", Training Accuracy= \" + \"{:.5f}\".format(acc)\n", + " step += 1\n", + " print \"Optimization Finished!\"\n", + " # Calculate accuracy for 256 mnist test images\n", + " print \"Testing Accuracy:\", sess.run(accuracy, feed_dict={x: mnist.test.images[:256], \n", + " y: mnist.test.labels[:256], \n", + " keep_prob: 1.})" + ] + } + ], + "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 +}