Skip to content

Instantly share code, notes, and snippets.

@Z30G0D

Z30G0D/Udacity deep learning - assignment 4

Created January 17, 2018 09:07
Show Gist options
  • Save Z30G0D/cf1ba895ae41605438b266a4fed01125 to your computer and use it in GitHub Desktop.
Save Z30G0D/cf1ba895ae41605438b266a4fed01125 to your computer and use it in GitHub Desktop.
Download ZIP
Fourth assignment of Udacity Deep learning course
Raw
Udacity deep learning - assignment 4
{
"cells": [
{
"cell_type": "markdown",
"metadata": {
"colab_type": "text",
"id": "4embtkV0pNxM"
},
"source": [
"Deep Learning\n",
"=============\n",
"\n",
"Assignment 4\n",
"------------\n",
"\n",
"Previously in `2_fullyconnected.ipynb` and `3_regularization.ipynb`, we trained fully connected networks to classify [notMNIST](http://yaroslavvb.blogspot.com/2011/09/notmnist-dataset.html) characters.\n",
"\n",
"The goal of this assignment is make the neural network convolutional."
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {
"cellView": "both",
"colab": {
"autoexec": {
"startup": false,
"wait_interval": 0
}
},
"colab_type": "code",
"id": "tm2CQN_Cpwj0"
},
"outputs": [],
"source": [
"# These are all the modules we'll be using later. Make sure you can import them\n",
"# before proceeding further.\n",
"from __future__ import print_function\n",
"import numpy as np\n",
"import tensorflow as tf\n",
"from six.moves import cPickle as pickle\n",
"from six.moves import range"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {
"cellView": "both",
"colab": {
"autoexec": {
"startup": false,
"wait_interval": 0
},
"output_extras": [
{
"item_id": 1
}
]
},
"colab_type": "code",
"executionInfo": {
"elapsed": 11948,
"status": "ok",
"timestamp": 1446658914837,
"user": {
"color": "",
"displayName": "",
"isAnonymous": false,
"isMe": true,
"permissionId": "",
"photoUrl": "",
"sessionId": "0",
"userId": ""
},
"user_tz": 480
},
"id": "y3-cj1bpmuxc",
"outputId": "016b1a51-0290-4b08-efdb-8c95ffc3cd01"
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Training set (2000, 28, 28) (2000,)\n",
"Validation set (100, 28, 28) (100,)\n",
"Test set (100, 28, 28) (100,)\n"
]
}
],
"source": [
"pickle_file = 'notMNIST.pickle'\n",
"\n",
"with open(pickle_file, 'rb') as f:\n",
" save = pickle.load(f, encoding='latin1')\n",
" train_dataset = save['train_dataset']\n",
" train_labels = save['train_labels']\n",
" valid_dataset = save['valid_dataset']\n",
" valid_labels = save['valid_labels']\n",
" test_dataset = save['test_dataset']\n",
" test_labels = save['test_labels']\n",
" del save # hint to help gc free up memory\n",
" print('Training set', train_dataset.shape, train_labels.shape)\n",
" print('Validation set', valid_dataset.shape, valid_labels.shape)\n",
" print('Test set', test_dataset.shape, test_labels.shape)"
]
},
{
"cell_type": "markdown",
"metadata": {
"colab_type": "text",
"id": "L7aHrm6nGDMB"
},
"source": [
"Reformat into a TensorFlow-friendly shape:\n",
"- convolutions need the image data formatted as a cube (width by height by #channels)\n",
"- labels as float 1-hot encodings."
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {
"cellView": "both",
"colab": {
"autoexec": {
"startup": false,
"wait_interval": 0
},
"output_extras": [
{
"item_id": 1
}
]
},
"colab_type": "code",
"executionInfo": {
"elapsed": 11952,
"status": "ok",
"timestamp": 1446658914857,
"user": {
"color": "",
"displayName": "",
"isAnonymous": false,
"isMe": true,
"permissionId": "",
"photoUrl": "",
"sessionId": "0",
"userId": ""
},
"user_tz": 480
},
"id": "IRSyYiIIGIzS",
"outputId": "650a208c-8359-4852-f4f5-8bf10e80ef6c"
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Training set (2000, 28, 28, 1) (2000, 10)\n",
"Validation set (100, 28, 28, 1) (100, 10)\n",
"Test set (100, 28, 28, 1) (100, 10)\n"
]
}
],
"source": [
"image_size = 28\n",
"num_labels = 10\n",
"num_channels = 1 # grayscale\n",
"\n",
"import numpy as np\n",
"\n",
"def reformat(dataset, labels):\n",
" dataset = dataset.reshape((-1, image_size, image_size, num_channels)).astype(np.float32)\n",
" labels = (np.arange(num_labels) == labels[:,None]).astype(np.float32)\n",
" return dataset, labels\n",
"train_dataset, train_labels = reformat(train_dataset, train_labels)\n",
"valid_dataset, valid_labels = reformat(valid_dataset, valid_labels)\n",
"test_dataset, test_labels = reformat(test_dataset, test_labels)\n",
"print('Training set', train_dataset.shape, train_labels.shape)\n",
"print('Validation set', valid_dataset.shape, valid_labels.shape)\n",
"print('Test set', test_dataset.shape, test_labels.shape)"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {
"cellView": "both",
"colab": {
"autoexec": {
"startup": false,
"wait_interval": 0
}
},
"colab_type": "code",
"id": "AgQDIREv02p1"
},
"outputs": [],
"source": [
"def accuracy(predictions, labels):\n",
" return (100.0 * np.sum(np.argmax(predictions, 1) == np.argmax(labels, 1))\n",
" / predictions.shape[0])"
]
},
{
"cell_type": "markdown",
"metadata": {
"colab_type": "text",
"id": "5rhgjmROXu2O"
},
"source": [
"Let's build a small network with two convolutional layers, followed by one fully connected layer. Convolutional networks are more expensive computationally, so we'll limit its depth and number of fully connected nodes. "
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {
"cellView": "both",
"colab": {
"autoexec": {
"startup": false,
"wait_interval": 0
}
},
"colab_type": "code",
"id": "IZYv70SvvOan"
},
"outputs": [],
"source": [
"batch_size = 16\n",
"patch_size = 5\n",
"depth = 16\n",
"num_hidden = 64\n",
"\n",
"graph = tf.Graph()\n",
"\n",
"with graph.as_default():\n",
"\n",
" # Input data.\n",
" tf_train_dataset = tf.placeholder(\n",
" tf.float32, shape=(batch_size, image_size, image_size, num_channels))\n",
" tf_train_labels = tf.placeholder(tf.float32, shape=(batch_size, num_labels))\n",
" tf_valid_dataset = tf.constant(valid_dataset)\n",
" tf_test_dataset = tf.constant(test_dataset)\n",
" \n",
" # Variables.\n",
" global_step = tf.Variable(0, trainable=False)\n",
" layer1_weights = tf.Variable(tf.truncated_normal(\n",
" [patch_size, patch_size, num_channels, depth], stddev=0.1))\n",
" layer1_biases = tf.Variable(tf.zeros([depth]))\n",
" layer2_weights = tf.Variable(tf.truncated_normal(\n",
" [patch_size, patch_size, depth, depth], stddev=0.1))\n",
" layer2_biases = tf.Variable(tf.constant(1.0, shape=[depth]))\n",
" layer3_weights = tf.Variable(tf.truncated_normal(\n",
" [image_size // 4 * image_size // 4 * depth, num_hidden], stddev=0.1))\n",
" layer3_biases = tf.Variable(tf.constant(1.0, shape=[num_hidden]))\n",
" layer4_weights = tf.Variable(tf.truncated_normal(\n",
" [num_hidden, num_labels], stddev=0.1))\n",
" layer4_biases = tf.Variable(tf.constant(1.0, shape=[num_labels]))\n",
" \n",
" # Model.\n",
" def model(data):\n",
" conv = tf.nn.conv2d(data, layer1_weights, [1, 2, 2, 1], padding='SAME')\n",
" hidden = tf.nn.relu(conv + layer1_biases)\n",
" conv = tf.nn.conv2d(hidden, layer2_weights, [1, 2, 2, 1], padding='SAME')\n",
" hidden = tf.nn.relu(conv + layer2_biases)\n",
" shape = hidden.get_shape().as_list()\n",
" reshape = tf.reshape(hidden, [shape[0], shape[1] * shape[2] * shape[3]])\n",
" hidden = tf.nn.relu(tf.matmul(reshape, layer3_weights) + layer3_biases)\n",
" return tf.matmul(hidden, layer4_weights) + layer4_biases\n",
" \n",
" # Training computation.\n",
" logits = model(tf_train_dataset)\n",
" loss = tf.reduce_mean(\n",
" tf.nn.softmax_cross_entropy_with_logits(labels=tf_train_labels, logits=logits))\n",
" \n",
" # Optimizer.\n",
" optimizer = tf.train.GradientDescentOptimizer(0.05).minimize(loss)\n",
" \n",
" # Predictions for the training, validation, and test data.\n",
" train_prediction = tf.nn.softmax(logits)\n",
" valid_prediction = tf.nn.softmax(model(tf_valid_dataset))\n",
" test_prediction = tf.nn.softmax(model(tf_test_dataset))"
]
},
{
"cell_type": "code",
"execution_count": 7,
"metadata": {
"cellView": "both",
"colab": {
"autoexec": {
"startup": false,
"wait_interval": 0
},
"output_extras": [
{
"item_id": 37
}
]
},
"colab_type": "code",
"executionInfo": {
"elapsed": 63292,
"status": "ok",
"timestamp": 1446658966251,
"user": {
"color": "",
"displayName": "",
"isAnonymous": false,
"isMe": true,
"permissionId": "",
"photoUrl": "",
"sessionId": "0",
"userId": ""
},
"user_tz": 480
},
"id": "noKFb2UovVFR",
"outputId": "28941338-2ef9-4088-8bd1-44295661e628"
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Initialized\n",
"Minibatch loss at step 0: 3.511478\n",
"Minibatch accuracy: 12.5%\n",
"Validation accuracy: 10.0%\n",
"Minibatch loss at step 50: 1.389899\n",
"Minibatch accuracy: 56.2%\n",
"Validation accuracy: 54.0%\n",
"Minibatch loss at step 100: 0.954439\n",
"Minibatch accuracy: 62.5%\n",
"Validation accuracy: 64.0%\n",
"Minibatch loss at step 150: 0.672420\n",
"Minibatch accuracy: 81.2%\n",
"Validation accuracy: 68.0%\n",
"Minibatch loss at step 200: 0.591480\n",
"Minibatch accuracy: 87.5%\n",
"Validation accuracy: 69.0%\n",
"Minibatch loss at step 250: 0.485368\n",
"Minibatch accuracy: 87.5%\n",
"Validation accuracy: 68.0%\n",
"Minibatch loss at step 300: 1.236439\n",
"Minibatch accuracy: 75.0%\n",
"Validation accuracy: 71.0%\n",
"Minibatch loss at step 350: 0.643501\n",
"Minibatch accuracy: 87.5%\n",
"Validation accuracy: 72.0%\n",
"Minibatch loss at step 400: 0.500857\n",
"Minibatch accuracy: 75.0%\n",
"Validation accuracy: 73.0%\n",
"Minibatch loss at step 450: 0.978600\n",
"Minibatch accuracy: 68.8%\n",
"Validation accuracy: 72.0%\n",
"Minibatch loss at step 500: 0.280913\n",
"Minibatch accuracy: 100.0%\n",
"Validation accuracy: 73.0%\n",
"Minibatch loss at step 550: 0.477067\n",
"Minibatch accuracy: 87.5%\n",
"Validation accuracy: 73.0%\n",
"Minibatch loss at step 600: 1.169504\n",
"Minibatch accuracy: 68.8%\n",
"Validation accuracy: 70.0%\n",
"Minibatch loss at step 650: 0.391893\n",
"Minibatch accuracy: 81.2%\n",
"Validation accuracy: 71.0%\n",
"Minibatch loss at step 700: 0.617436\n",
"Minibatch accuracy: 81.2%\n",
"Validation accuracy: 74.0%\n",
"Minibatch loss at step 750: 0.412628\n",
"Minibatch accuracy: 81.2%\n",
"Validation accuracy: 75.0%\n",
"Minibatch loss at step 800: 0.696503\n",
"Minibatch accuracy: 81.2%\n",
"Validation accuracy: 73.0%\n",
"Minibatch loss at step 850: 0.206645\n",
"Minibatch accuracy: 100.0%\n",
"Validation accuracy: 73.0%\n",
"Minibatch loss at step 900: 0.177372\n",
"Minibatch accuracy: 93.8%\n",
"Validation accuracy: 73.0%\n",
"Minibatch loss at step 950: 0.265233\n",
"Minibatch accuracy: 93.8%\n",
"Validation accuracy: 75.0%\n",
"Minibatch loss at step 1000: 0.046155\n",
"Minibatch accuracy: 100.0%\n",
"Validation accuracy: 75.0%\n",
"Test accuracy: 80.0%\n"
]
}
],
"source": [
"num_steps = 1001\n",
"\n",
"with tf.Session(graph=graph) as session:\n",
" tf.global_variables_initializer().run()\n",
" print('Initialized')\n",
" for step in range(num_steps):\n",
" offset = (step * batch_size) % (train_labels.shape[0] - batch_size)\n",
" batch_data = train_dataset[offset:(offset + batch_size), :, :, :]\n",
" batch_labels = train_labels[offset:(offset + batch_size), :]\n",
" feed_dict = {tf_train_dataset : batch_data, tf_train_labels : batch_labels}\n",
" _, l, predictions = session.run(\n",
" [optimizer, loss, train_prediction], feed_dict=feed_dict)\n",
" if (step % 50 == 0):\n",
" print('Minibatch loss at step %d: %f' % (step, l))\n",
" print('Minibatch accuracy: %.1f%%' % accuracy(predictions, batch_labels))\n",
" print('Validation accuracy: %.1f%%' % accuracy(\n",
" valid_prediction.eval(), valid_labels))\n",
" print('Test accuracy: %.1f%%' % accuracy(test_prediction.eval(), test_labels))"
]
},
{
"cell_type": "markdown",
"metadata": {
"colab_type": "text",
"id": "KedKkn4EutIK"
},
"source": [
"---\n",
"Problem 1\n",
"---------\n",
"\n",
"The convolutional model above uses convolutions with stride 2 to reduce the dimensionality. Replace the strides by a max pooling operation (`nn.max_pool()`) of stride 2 and kernel size 2.\n",
"\n",
"---"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Ok, we are asked to introduce a max pooling operation, this operation will take the highest reaction in the vicinity of the pixel and implement it in the new feature map, hence subsampling our data."
]
},
{
"cell_type": "code",
"execution_count": 8,
"metadata": {},
"outputs": [],
"source": [
"batch_size = 16\n",
"patch_size = 5\n",
"depth = 16\n",
"num_hidden = 64\n",
"\n",
"graph = tf.Graph()\n",
"\n",
"with graph.as_default():\n",
"\n",
" # Input data.\n",
" tf_train_dataset = tf.placeholder(\n",
" tf.float32, shape=(batch_size, image_size, image_size, num_channels))\n",
" tf_train_labels = tf.placeholder(tf.float32, shape=(batch_size, num_labels))\n",
" tf_valid_dataset = tf.constant(valid_dataset)\n",
" tf_test_dataset = tf.constant(test_dataset)\n",
" \n",
" # Variables.\n",
" layer1_weights = tf.Variable(tf.truncated_normal(\n",
" [patch_size, patch_size, num_channels, depth], stddev=0.1))\n",
" layer1_biases = tf.Variable(tf.zeros([depth]))\n",
" layer2_weights = tf.Variable(tf.truncated_normal(\n",
" [patch_size, patch_size, depth, depth], stddev=0.1))\n",
" layer2_biases = tf.Variable(tf.constant(1.0, shape=[depth]))\n",
" layer3_weights = tf.Variable(tf.truncated_normal(\n",
" [image_size // 4 * image_size // 4 * depth, num_hidden], stddev=0.1))\n",
" layer3_biases = tf.Variable(tf.constant(1.0, shape=[num_hidden]))\n",
" layer4_weights = tf.Variable(tf.truncated_normal(\n",
" [num_hidden, num_labels], stddev=0.1))\n",
" layer4_biases = tf.Variable(tf.constant(1.0, shape=[num_labels]))\n",
" \n",
" # Model.\n",
" def model(data):\n",
" # reduce convolution to 1 stride\n",
" conv = tf.nn.conv2d(data, layer1_weights, [1, 1, 1, 1], padding='SAME')\n",
" #add pooling, kernel = 2, stride = 2\n",
" pool1 = tf.nn.max_pool(conv, [1, 2, 2, 1], [1, 2, 2, 1], padding='SAME')\n",
" hidden = tf.nn.relu(pool1 + layer1_biases)\n",
" conv = tf.nn.conv2d(hidden, layer2_weights, [1, 1, 1, 1], padding='SAME')\n",
" pool2 = tf.nn.max_pool(conv, [1, 2, 2, 1], [1, 2, 2, 1], padding='SAME')\n",
" hidden = tf.nn.relu(pool2 + layer2_biases)\n",
" shape = hidden.get_shape().as_list()\n",
" reshape = tf.reshape(hidden, [shape[0], shape[1] * shape[2] * shape[3]])\n",
" hidden = tf.nn.relu(tf.matmul(reshape, layer3_weights) + layer3_biases)\n",
" return tf.matmul(hidden, layer4_weights) + layer4_biases\n",
" \n",
" # Training computation.\n",
" logits = model(tf_train_dataset)\n",
" loss = tf.reduce_mean(\n",
" tf.nn.softmax_cross_entropy_with_logits(labels=tf_train_labels, logits=logits))\n",
" \n",
" # Optimizer.\n",
" optimizer = tf.train.GradientDescentOptimizer(0.05).minimize(loss)\n",
" \n",
" # Predictions for the training, validation, and test data.\n",
" train_prediction = tf.nn.softmax(logits)\n",
" valid_prediction = tf.nn.softmax(model(tf_valid_dataset))\n",
" test_prediction = tf.nn.softmax(model(tf_test_dataset))"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Let's run the graph..."
]
},
{
"cell_type": "code",
"execution_count": 9,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Initialized\n",
"Minibatch loss at step 0: 4.143212\n",
"Minibatch accuracy: 6.2%\n",
"Validation accuracy: 10.0%\n",
"Minibatch loss at step 50: 1.858743\n",
"Minibatch accuracy: 43.8%\n",
"Validation accuracy: 36.0%\n",
"Minibatch loss at step 100: 1.196889\n",
"Minibatch accuracy: 50.0%\n",
"Validation accuracy: 59.0%\n",
"Minibatch loss at step 150: 0.710960\n",
"Minibatch accuracy: 75.0%\n",
"Validation accuracy: 69.0%\n",
"Minibatch loss at step 200: 0.591121\n",
"Minibatch accuracy: 81.2%\n",
"Validation accuracy: 69.0%\n",
"Minibatch loss at step 250: 0.364006\n",
"Minibatch accuracy: 87.5%\n",
"Validation accuracy: 76.0%\n",
"Minibatch loss at step 300: 1.070594\n",
"Minibatch accuracy: 75.0%\n",
"Validation accuracy: 73.0%\n",
"Minibatch loss at step 350: 0.787328\n",
"Minibatch accuracy: 81.2%\n",
"Validation accuracy: 75.0%\n",
"Minibatch loss at step 400: 0.453832\n",
"Minibatch accuracy: 87.5%\n",
"Validation accuracy: 74.0%\n",
"Minibatch loss at step 450: 1.025140\n",
"Minibatch accuracy: 81.2%\n",
"Validation accuracy: 76.0%\n",
"Minibatch loss at step 500: 0.279667\n",
"Minibatch accuracy: 93.8%\n",
"Validation accuracy: 76.0%\n",
"Minibatch loss at step 550: 0.498919\n",
"Minibatch accuracy: 87.5%\n",
"Validation accuracy: 76.0%\n",
"Minibatch loss at step 600: 0.709353\n",
"Minibatch accuracy: 75.0%\n",
"Validation accuracy: 79.0%\n",
"Minibatch loss at step 650: 0.353762\n",
"Minibatch accuracy: 87.5%\n",
"Validation accuracy: 71.0%\n",
"Minibatch loss at step 700: 0.706299\n",
"Minibatch accuracy: 81.2%\n",
"Validation accuracy: 75.0%\n",
"Minibatch loss at step 750: 0.397130\n",
"Minibatch accuracy: 81.2%\n",
"Validation accuracy: 75.0%\n",
"Minibatch loss at step 800: 0.462259\n",
"Minibatch accuracy: 87.5%\n",
"Validation accuracy: 74.0%\n",
"Minibatch loss at step 850: 0.137734\n",
"Minibatch accuracy: 100.0%\n",
"Validation accuracy: 76.0%\n",
"Minibatch loss at step 900: 0.162059\n",
"Minibatch accuracy: 93.8%\n",
"Validation accuracy: 72.0%\n",
"Minibatch loss at step 950: 0.254761\n",
"Minibatch accuracy: 93.8%\n",
"Validation accuracy: 77.0%\n",
"Minibatch loss at step 1000: 0.079590\n",
"Minibatch accuracy: 100.0%\n",
"Validation accuracy: 73.0%\n",
"Test accuracy: 85.0%\n"
]
}
],
"source": [
"num_steps = 1001\n",
"\n",
"with tf.Session(graph=graph) as session:\n",
" tf.global_variables_initializer().run()\n",
" print('Initialized')\n",
" for step in range(num_steps):\n",
" offset = (step * batch_size) % (train_labels.shape[0] - batch_size)\n",
" batch_data = train_dataset[offset:(offset + batch_size), :, :, :]\n",
" batch_labels = train_labels[offset:(offset + batch_size), :]\n",
" feed_dict = {tf_train_dataset : batch_data, tf_train_labels : batch_labels}\n",
" _, l, predictions = session.run(\n",
" [optimizer, loss, train_prediction], feed_dict=feed_dict)\n",
" if (step % 50 == 0):\n",
" print('Minibatch loss at step %d: %f' % (step, l))\n",
" print('Minibatch accuracy: %.1f%%' % accuracy(predictions, batch_labels))\n",
" print('Validation accuracy: %.1f%%' % accuracy(\n",
" valid_prediction.eval(), valid_labels))\n",
" print('Test accuracy: %.1f%%' % accuracy(test_prediction.eval(), test_labels))"
]
},
{
"cell_type": "markdown",
"metadata": {
"colab_type": "text",
"id": "klf21gpbAgb-"
},
"source": [
"---\n",
"Problem 2\n",
"---------\n",
"\n",
"Try to get the best performance you can using a convolutional net. Look for example at the classic [LeNet5](http://yann.lecun.com/exdb/lenet/) architecture, adding Dropout, and/or adding learning rate decay.\n",
"\n",
"---"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Ok, since I have problems with processing at the moment and I still need to set up my AWS/ have my own GPU to run it on my personal server, I will introduce the learning rate decay "
]
},
{
"cell_type": "code",
"execution_count": 29,
"metadata": {},
"outputs": [],
"source": [
"batch_size = 16\n",
"patch_size = 5\n",
"depth = 16\n",
"num_hidden = 64\n",
"starter_learning_rate = 0.05\n",
"\n",
"graph = tf.Graph()\n",
"\n",
"with graph.as_default():\n",
"\n",
" # Input data.\n",
" tf_train_dataset = tf.placeholder(\n",
" tf.float32, shape=(batch_size, image_size, image_size, num_channels))\n",
" tf_train_labels = tf.placeholder(tf.float32, shape=(batch_size, num_labels))\n",
" tf_valid_dataset = tf.constant(valid_dataset)\n",
" tf_test_dataset = tf.constant(test_dataset)\n",
" \n",
" # Variables.\n",
" global_step = tf.Variable(0, trainable=False)\n",
" layer1_weights = tf.Variable(tf.truncated_normal(\n",
" [patch_size, patch_size, num_channels, depth], stddev=0.1))\n",
" layer1_biases = tf.Variable(tf.zeros([depth]))\n",
" layer2_weights = tf.Variable(tf.truncated_normal(\n",
" [patch_size, patch_size, depth, depth], stddev=0.1))\n",
" layer2_biases = tf.Variable(tf.constant(1.0, shape=[depth]))\n",
" layer3_weights = tf.Variable(tf.truncated_normal(\n",
" [image_size // 4 * image_size // 4 * depth, num_hidden], stddev=0.1))\n",
" layer3_biases = tf.Variable(tf.constant(1.0, shape=[num_hidden]))\n",
" layer4_weights = tf.Variable(tf.truncated_normal(\n",
" [num_hidden, num_labels], stddev=0.1))\n",
" layer4_biases = tf.Variable(tf.constant(1.0, shape=[num_labels]))\n",
" \n",
" # Model.\n",
" def model(data):\n",
" # reduce convolution to 1 stride\n",
" conv = tf.nn.conv2d(data, layer1_weights, [1, 1, 1, 1], padding='SAME')\n",
" #add pooling, kernel = 2, stride = 2\n",
" pool1 = tf.nn.max_pool(conv, [1, 2, 2, 1], [1, 2, 2, 1], padding='SAME')\n",
" hidden = tf.nn.relu(pool1 + layer1_biases)\n",
" conv = tf.nn.conv2d(hidden, layer2_weights, [1, 1, 1, 1], padding='SAME')\n",
" pool2 = tf.nn.max_pool(conv, [1, 2, 2, 1], [1, 2, 2, 1], padding='SAME')\n",
" hidden = tf.nn.relu(pool2 + layer2_biases)\n",
" shape = hidden.get_shape().as_list()\n",
" reshape = tf.reshape(hidden, [shape[0], shape[1] * shape[2] * shape[3]])\n",
" hidden = tf.nn.relu(tf.matmul(reshape, layer3_weights) + layer3_biases)\n",
" return tf.matmul(hidden, layer4_weights) + layer4_biases\n",
" \n",
" # Training computation.\n",
" logits = model(tf_train_dataset)\n",
" loss = tf.reduce_mean(\n",
" tf.nn.softmax_cross_entropy_with_logits(labels=tf_train_labels, logits=logits))\n",
" \n",
" # Optimizer.\n",
" learning_rate = tf.train.exponential_decay(starter_learning_rate, global_step, 2000, 0.95, staircase=True)\n",
" optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(loss ,global_step=global_step)\n",
" \n",
" # Predictions for the training, validation, and test data.\n",
" train_prediction = tf.nn.softmax(logits)\n",
" valid_prediction = tf.nn.softmax(model(tf_valid_dataset))\n",
" test_prediction = tf.nn.softmax(model(tf_test_dataset))"
]
},
{
"cell_type": "code",
"execution_count": 30,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Initialized\n",
"Minibatch loss at step 0: 3.491250\n",
"Minibatch accuracy: 18.8%\n",
"Validation accuracy: 12.0%\n",
"Minibatch loss at step 50: 1.075704\n",
"Minibatch accuracy: 62.5%\n",
"Validation accuracy: 65.0%\n",
"Minibatch loss at step 100: 0.905266\n",
"Minibatch accuracy: 62.5%\n",
"Validation accuracy: 66.0%\n",
"Minibatch loss at step 150: 0.670949\n",
"Minibatch accuracy: 81.2%\n",
"Validation accuracy: 70.0%\n",
"Minibatch loss at step 200: 0.464487\n",
"Minibatch accuracy: 87.5%\n",
"Validation accuracy: 70.0%\n",
"Minibatch loss at step 250: 0.358510\n",
"Minibatch accuracy: 87.5%\n",
"Validation accuracy: 72.0%\n",
"Minibatch loss at step 300: 0.868001\n",
"Minibatch accuracy: 75.0%\n",
"Validation accuracy: 72.0%\n",
"Minibatch loss at step 350: 0.489978\n",
"Minibatch accuracy: 75.0%\n",
"Validation accuracy: 75.0%\n",
"Minibatch loss at step 400: 0.350634\n",
"Minibatch accuracy: 93.8%\n",
"Validation accuracy: 72.0%\n",
"Minibatch loss at step 450: 1.272129\n",
"Minibatch accuracy: 75.0%\n",
"Validation accuracy: 72.0%\n",
"Minibatch loss at step 500: 0.228550\n",
"Minibatch accuracy: 93.8%\n",
"Validation accuracy: 72.0%\n",
"Minibatch loss at step 550: 0.308218\n",
"Minibatch accuracy: 87.5%\n",
"Validation accuracy: 74.0%\n",
"Minibatch loss at step 600: 0.819818\n",
"Minibatch accuracy: 75.0%\n",
"Validation accuracy: 73.0%\n",
"Minibatch loss at step 650: 0.431710\n",
"Minibatch accuracy: 87.5%\n",
"Validation accuracy: 68.0%\n",
"Minibatch loss at step 700: 0.814631\n",
"Minibatch accuracy: 81.2%\n",
"Validation accuracy: 73.0%\n",
"Minibatch loss at step 750: 0.527242\n",
"Minibatch accuracy: 87.5%\n",
"Validation accuracy: 73.0%\n",
"Minibatch loss at step 800: 0.616049\n",
"Minibatch accuracy: 81.2%\n",
"Validation accuracy: 75.0%\n",
"Minibatch loss at step 850: 0.123858\n",
"Minibatch accuracy: 93.8%\n",
"Validation accuracy: 73.0%\n",
"Minibatch loss at step 900: 0.093429\n",
"Minibatch accuracy: 100.0%\n",
"Validation accuracy: 72.0%\n",
"Minibatch loss at step 950: 0.201918\n",
"Minibatch accuracy: 93.8%\n",
"Validation accuracy: 72.0%\n",
"Minibatch loss at step 1000: 0.038134\n",
"Minibatch accuracy: 100.0%\n",
"Validation accuracy: 74.0%\n",
"Test accuracy: 88.0%\n"
]
}
],
"source": [
"num_steps = 1001\n",
"\n",
"with tf.Session(graph=graph) as session:\n",
" tf.global_variables_initializer().run()\n",
" print('Initialized')\n",
" for step in range(num_steps):\n",
" offset = (step * batch_size) % (train_labels.shape[0] - batch_size)\n",
" batch_data = train_dataset[offset:(offset + batch_size), :, :, :]\n",
" batch_labels = train_labels[offset:(offset + batch_size), :]\n",
" feed_dict = {tf_train_dataset : batch_data, tf_train_labels : batch_labels}\n",
" _, l, predictions = session.run(\n",
" [optimizer, loss, train_prediction], feed_dict=feed_dict)\n",
" if (step % 50 == 0):\n",
" print('Minibatch loss at step %d: %f' % (step, l))\n",
" print('Minibatch accuracy: %.1f%%' % accuracy(predictions, batch_labels))\n",
" print('Validation accuracy: %.1f%%' % accuracy(\n",
" valid_prediction.eval(), valid_labels))\n",
" print('Test accuracy: %.1f%%' % accuracy(test_prediction.eval(), test_labels))"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"We can see that the test accuracy went up from 85% to 88% by changing the learning rate to a decaying exponential learning rate"
]
}
],
"metadata": {
"colab": {
"default_view": {},
"name": "4_convolutions.ipynb",
"provenance": [],
"version": "0.3.2",
"views": {}
},
"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": "3.5.4"
}
},
"nbformat": 4,
"nbformat_minor": 1
}
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment