wehub-resource-sync
230 lines
7.8 KiB
Plaintext
230 lines
7.8 KiB
Plaintext
{
|
|
"cells": [
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"Deep Learning Models -- A collection of various deep learning architectures, models, and tips for TensorFlow and PyTorch in Jupyter Notebooks.\n",
|
|
"- Author: Sebastian Raschka\n",
|
|
"- GitHub Repository: https://github.com/rasbt/deeplearning-models"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": 1,
|
|
"metadata": {},
|
|
"outputs": [
|
|
{
|
|
"name": "stdout",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"Sebastian Raschka \n",
|
|
"\n",
|
|
"CPython 3.6.1\n",
|
|
"IPython 6.0.0\n",
|
|
"\n",
|
|
"tensorflow 1.2.0\n"
|
|
]
|
|
}
|
|
],
|
|
"source": [
|
|
"%load_ext watermark\n",
|
|
"%watermark -a 'Sebastian Raschka' -v -p tensorflow"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"# Model Zoo -- Multilayer Perceptron with Batch Normalization"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": 2,
|
|
"metadata": {},
|
|
"outputs": [
|
|
{
|
|
"name": "stdout",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"Extracting ./train-images-idx3-ubyte.gz\n",
|
|
"Extracting ./train-labels-idx1-ubyte.gz\n",
|
|
"Extracting ./t10k-images-idx3-ubyte.gz\n",
|
|
"Extracting ./t10k-labels-idx1-ubyte.gz\n"
|
|
]
|
|
}
|
|
],
|
|
"source": [
|
|
"import tensorflow as tf\n",
|
|
"from tensorflow.examples.tutorials.mnist import input_data\n",
|
|
"\n",
|
|
"\n",
|
|
"##########################\n",
|
|
"### DATASET\n",
|
|
"##########################\n",
|
|
"\n",
|
|
"mnist = input_data.read_data_sets(\"./\", one_hot=True)\n",
|
|
"\n",
|
|
"\n",
|
|
"##########################\n",
|
|
"### SETTINGS\n",
|
|
"##########################\n",
|
|
"\n",
|
|
"# Hyperparameters\n",
|
|
"learning_rate = 0.1\n",
|
|
"training_epochs = 10\n",
|
|
"batch_size = 64\n",
|
|
"\n",
|
|
"# Architecture\n",
|
|
"n_hidden_1 = 128\n",
|
|
"n_hidden_2 = 256\n",
|
|
"n_input = 784\n",
|
|
"n_classes = 10\n",
|
|
"\n",
|
|
"# Other\n",
|
|
"random_seed = 123\n",
|
|
"\n",
|
|
"\n",
|
|
"##########################\n",
|
|
"### GRAPH DEFINITION\n",
|
|
"##########################\n",
|
|
"\n",
|
|
"g = tf.Graph()\n",
|
|
"with g.as_default():\n",
|
|
" \n",
|
|
" tf.set_random_seed(random_seed)\n",
|
|
" \n",
|
|
" # Batchnorm settings\n",
|
|
" training_phase = tf.placeholder(tf.bool, None, name='training_phase')\n",
|
|
"\n",
|
|
" # Input data\n",
|
|
" tf_x = tf.placeholder(tf.float32, [None, n_input], name='features')\n",
|
|
" tf_y = tf.placeholder(tf.float32, [None, n_classes], name='targets')\n",
|
|
"\n",
|
|
" # Multilayer perceptron\n",
|
|
" layer_1 = tf.layers.dense(tf_x, n_hidden_1, \n",
|
|
" activation=None, # Batchnorm comes before nonlinear activation\n",
|
|
" use_bias=False, # Note that no bias unit is used in batchnorm\n",
|
|
" kernel_initializer=tf.truncated_normal_initializer(stddev=0.1))\n",
|
|
" \n",
|
|
" layer_1 = tf.layers.batch_normalization(layer_1, training=training_phase)\n",
|
|
" layer_1 = tf.nn.relu(layer_1)\n",
|
|
" \n",
|
|
" layer_2 = tf.layers.dense(layer_1, n_hidden_2, \n",
|
|
" activation=None,\n",
|
|
" use_bias=False,\n",
|
|
" kernel_initializer=tf.truncated_normal_initializer(stddev=0.1))\n",
|
|
" layer_2 = tf.layers.batch_normalization(layer_2, training=training_phase)\n",
|
|
" layer_2 = tf.nn.relu(layer_2)\n",
|
|
" \n",
|
|
" out_layer = tf.layers.dense(layer_2, n_classes, activation=None, name='logits')\n",
|
|
"\n",
|
|
" # Loss and optimizer\n",
|
|
" loss = tf.nn.softmax_cross_entropy_with_logits(logits=out_layer, labels=tf_y)\n",
|
|
" cost = tf.reduce_mean(loss, name='cost')\n",
|
|
" \n",
|
|
" # control dependency to ensure that batchnorm parameters are also updated\n",
|
|
" with tf.control_dependencies(tf.get_collection(tf.GraphKeys.UPDATE_OPS)):\n",
|
|
" optimizer = tf.train.GradientDescentOptimizer(learning_rate=learning_rate)\n",
|
|
" train = optimizer.minimize(cost, name='train')\n",
|
|
"\n",
|
|
" # Prediction\n",
|
|
" correct_prediction = tf.equal(tf.argmax(tf_y, 1), tf.argmax(out_layer, 1))\n",
|
|
" accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32), name='accuracy')"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": 3,
|
|
"metadata": {},
|
|
"outputs": [
|
|
{
|
|
"name": "stdout",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"Epoch: 001 | AvgCost: 0.280 | Train/Valid ACC: 0.962/0.960\n",
|
|
"Epoch: 002 | AvgCost: 0.131 | Train/Valid ACC: 0.978/0.972\n",
|
|
"Epoch: 003 | AvgCost: 0.095 | Train/Valid ACC: 0.984/0.973\n",
|
|
"Epoch: 004 | AvgCost: 0.074 | Train/Valid ACC: 0.988/0.976\n",
|
|
"Epoch: 005 | AvgCost: 0.059 | Train/Valid ACC: 0.992/0.980\n",
|
|
"Epoch: 006 | AvgCost: 0.049 | Train/Valid ACC: 0.995/0.980\n",
|
|
"Epoch: 007 | AvgCost: 0.039 | Train/Valid ACC: 0.996/0.979\n",
|
|
"Epoch: 008 | AvgCost: 0.033 | Train/Valid ACC: 0.997/0.981\n",
|
|
"Epoch: 009 | AvgCost: 0.030 | Train/Valid ACC: 0.997/0.977\n",
|
|
"Epoch: 010 | AvgCost: 0.024 | Train/Valid ACC: 0.998/0.979\n",
|
|
"Test ACC: 0.977\n"
|
|
]
|
|
}
|
|
],
|
|
"source": [
|
|
"import numpy as np\n",
|
|
"\n",
|
|
"##########################\n",
|
|
"### TRAINING & EVALUATION\n",
|
|
"##########################\n",
|
|
" \n",
|
|
"with tf.Session(graph=g) as sess:\n",
|
|
" sess.run(tf.global_variables_initializer())\n",
|
|
"\n",
|
|
" np.random.seed(random_seed) # random seed for mnist iterator\n",
|
|
" for epoch in range(training_epochs):\n",
|
|
" avg_cost = 0.\n",
|
|
" total_batch = mnist.train.num_examples // batch_size\n",
|
|
"\n",
|
|
" for i in range(total_batch):\n",
|
|
" batch_x, batch_y = mnist.train.next_batch(batch_size)\n",
|
|
" _, c = sess.run(['train', 'cost:0'], feed_dict={'features:0': batch_x,\n",
|
|
" 'targets:0': batch_y,\n",
|
|
" 'training_phase:0': True})\n",
|
|
" avg_cost += c\n",
|
|
" \n",
|
|
" train_acc = sess.run('accuracy:0', feed_dict={'features:0': mnist.train.images,\n",
|
|
" 'targets:0': mnist.train.labels,\n",
|
|
" 'training_phase:0': False})\n",
|
|
" valid_acc = sess.run('accuracy:0', feed_dict={'features:0': mnist.validation.images,\n",
|
|
" 'targets:0': mnist.validation.labels,\n",
|
|
" 'training_phase:0': False}) \n",
|
|
" \n",
|
|
" print(\"Epoch: %03d | AvgCost: %.3f\" % (epoch + 1, avg_cost / (i + 1)), end=\"\")\n",
|
|
" print(\" | Train/Valid ACC: %.3f/%.3f\" % (train_acc, valid_acc))\n",
|
|
" \n",
|
|
" test_acc = sess.run('accuracy:0', feed_dict={'features:0': mnist.test.images,\n",
|
|
" 'targets:0': mnist.test.labels,\n",
|
|
" 'training_phase:0': False})\n",
|
|
" print('Test ACC: %.3f' % test_acc)"
|
|
]
|
|
},
|
|
{
|
|
"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": "3.7.1"
|
|
}
|
|
},
|
|
"nbformat": 4,
|
|
"nbformat_minor": 2
|
|
}
|