Bi-directional Recurrent Neural Network Example¶
Build a bi-directional recurrent neural network (LSTM) with TensorFlow.
- Author: Aymeric Damien
- Project: https://github.com/aymericdamien/TensorFlow-Examples/
BiRNN Overview¶
References:
- Long Short Term Memory, Sepp Hochreiter & Jurgen Schmidhuber, Neural Computation 9(8): 1735-1780, 1997.
MNIST Dataset Overview¶
This example is using MNIST handwritten digits. The dataset contains 60,000 examples for training and 10,000 examples for testing. The digits have been size-normalized and centered in a fixed-size image (28x28 pixels) with values from 0 to 1. For simplicity, each image has been flattened and converted to a 1-D numpy array of 784 features (28*28).
To classify images using a recurrent neural network, we consider every image row as a sequence of pixels. Because MNIST image shape is 28*28px, we will then handle 28 sequences of 28 timesteps for every sample.
More info: http://yann.lecun.com/exdb/mnist/
In [1]:
from __future__ import print_function
import tensorflow as tf
from tensorflow.contrib import rnn
import numpy as np
# Import MNIST data
from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets("/tmp/data/", one_hot=True)
Extracting /tmp/data/train-images-idx3-ubyte.gz Extracting /tmp/data/train-labels-idx1-ubyte.gz Extracting /tmp/data/t10k-images-idx3-ubyte.gz Extracting /tmp/data/t10k-labels-idx1-ubyte.gz
In [2]:
# Training Parameters
learning_rate = 0.001
training_steps = 10000
batch_size = 128
display_step = 200
# Network Parameters
num_input = 28 # MNIST data input (img shape: 28*28)
timesteps = 28 # timesteps
num_hidden = 128 # hidden layer num of features
num_classes = 10 # MNIST total classes (0-9 digits)
# tf Graph input
X = tf.placeholder("float", [None, timesteps, num_input])
Y = tf.placeholder("float", [None, num_classes])
In [3]:
# Define weights
weights = {
# Hidden layer weights => 2*n_hidden because of forward + backward cells
'out': tf.Variable(tf.random_normal([2*num_hidden, num_classes]))
}
biases = {
'out': tf.Variable(tf.random_normal([num_classes]))
}
In [4]:
def BiRNN(x, weights, biases):
# Prepare data shape to match `rnn` function requirements
# Current data input shape: (batch_size, timesteps, n_input)
# Required shape: 'timesteps' tensors list of shape (batch_size, num_input)
# Unstack to get a list of 'timesteps' tensors of shape (batch_size, num_input)
x = tf.unstack(x, timesteps, 1)
# Define lstm cells with tensorflow
# Forward direction cell
lstm_fw_cell = rnn.BasicLSTMCell(num_hidden, forget_bias=1.0)
# Backward direction cell
lstm_bw_cell = rnn.BasicLSTMCell(num_hidden, forget_bias=1.0)
# Get lstm cell output
try:
outputs, _, _ = rnn.static_bidirectional_rnn(lstm_fw_cell, lstm_bw_cell, x,
dtype=tf.float32)
except Exception: # Old TensorFlow version only returns outputs not states
outputs = rnn.static_bidirectional_rnn(lstm_fw_cell, lstm_bw_cell, x,
dtype=tf.float32)
# Linear activation, using rnn inner loop last output
return tf.matmul(outputs[-1], weights['out']) + biases['out']
In [5]:
logits = BiRNN(X, weights, biases)
prediction = tf.nn.softmax(logits)
# Define loss and optimizer
loss_op = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(
logits=logits, labels=Y))
optimizer = tf.train.GradientDescentOptimizer(learning_rate=learning_rate)
train_op = optimizer.minimize(loss_op)
# Evaluate model (with test logits, for dropout to be disabled)
correct_pred = tf.equal(tf.argmax(prediction, 1), tf.argmax(Y, 1))
accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))
# Initialize the variables (i.e. assign their default value)
init = tf.global_variables_initializer()
In [6]:
# Start training
with tf.Session() as sess:
# Run the initializer
sess.run(init)
for step in range(1, training_steps+1):
batch_x, batch_y = mnist.train.next_batch(batch_size)
# Reshape data to get 28 seq of 28 elements
batch_x = batch_x.reshape((batch_size, timesteps, num_input))
# Run optimization op (backprop)
sess.run(train_op, feed_dict={X: batch_x, Y: batch_y})
if step % display_step == 0 or step == 1:
# Calculate batch loss and accuracy
loss, acc = sess.run([loss_op, accuracy], feed_dict={X: batch_x,
Y: batch_y})
print("Step " + str(step) + ", Minibatch Loss= " + \
"{:.4f}".format(loss) + ", Training Accuracy= " + \
"{:.3f}".format(acc))
print("Optimization Finished!")
# Calculate accuracy for 128 mnist test images
test_len = 128
test_data = mnist.test.images[:test_len].reshape((-1, timesteps, num_input))
test_label = mnist.test.labels[:test_len]
print("Testing Accuracy:", \
sess.run(accuracy, feed_dict={X: test_data, Y: test_label}))
Step 1, Minibatch Loss= 2.6218, Training Accuracy= 0.086 Step 200, Minibatch Loss= 2.1900, Training Accuracy= 0.211 Step 400, Minibatch Loss= 2.0144, Training Accuracy= 0.375 Step 600, Minibatch Loss= 1.8729, Training Accuracy= 0.445 Step 800, Minibatch Loss= 1.8000, Training Accuracy= 0.469 Step 1000, Minibatch Loss= 1.7244, Training Accuracy= 0.453 Step 1200, Minibatch Loss= 1.5657, Training Accuracy= 0.523 Step 1400, Minibatch Loss= 1.5473, Training Accuracy= 0.547 Step 1600, Minibatch Loss= 1.5288, Training Accuracy= 0.500 Step 1800, Minibatch Loss= 1.4203, Training Accuracy= 0.555 Step 2000, Minibatch Loss= 1.2525, Training Accuracy= 0.641 Step 2200, Minibatch Loss= 1.2696, Training Accuracy= 0.594 Step 2400, Minibatch Loss= 1.2000, Training Accuracy= 0.664 Step 2600, Minibatch Loss= 1.1017, Training Accuracy= 0.625 Step 2800, Minibatch Loss= 1.2656, Training Accuracy= 0.578 Step 3000, Minibatch Loss= 1.0830, Training Accuracy= 0.656 Step 3200, Minibatch Loss= 1.1522, Training Accuracy= 0.633 Step 3400, Minibatch Loss= 0.9484, Training Accuracy= 0.680 Step 3600, Minibatch Loss= 1.0470, Training Accuracy= 0.641 Step 3800, Minibatch Loss= 1.0609, Training Accuracy= 0.586 Step 4000, Minibatch Loss= 1.1853, Training Accuracy= 0.648 Step 4200, Minibatch Loss= 0.9438, Training Accuracy= 0.750 Step 4400, Minibatch Loss= 0.7986, Training Accuracy= 0.766 Step 4600, Minibatch Loss= 0.8070, Training Accuracy= 0.750 Step 4800, Minibatch Loss= 0.8382, Training Accuracy= 0.734 Step 5000, Minibatch Loss= 0.7397, Training Accuracy= 0.766 Step 5200, Minibatch Loss= 0.7870, Training Accuracy= 0.727 Step 5400, Minibatch Loss= 0.6380, Training Accuracy= 0.828 Step 5600, Minibatch Loss= 0.7975, Training Accuracy= 0.719 Step 5800, Minibatch Loss= 0.7934, Training Accuracy= 0.766 Step 6000, Minibatch Loss= 0.6628, Training Accuracy= 0.805 Step 6200, Minibatch Loss= 0.7958, Training Accuracy= 0.672 Step 6400, Minibatch Loss= 0.6582, Training Accuracy= 0.773 Step 6600, Minibatch Loss= 0.5908, Training Accuracy= 0.812 Step 6800, Minibatch Loss= 0.6182, Training Accuracy= 0.820 Step 7000, Minibatch Loss= 0.5513, Training Accuracy= 0.812 Step 7200, Minibatch Loss= 0.6683, Training Accuracy= 0.789 Step 7400, Minibatch Loss= 0.5337, Training Accuracy= 0.828 Step 7600, Minibatch Loss= 0.6428, Training Accuracy= 0.805 Step 7800, Minibatch Loss= 0.6708, Training Accuracy= 0.797 Step 8000, Minibatch Loss= 0.4664, Training Accuracy= 0.852 Step 8200, Minibatch Loss= 0.4249, Training Accuracy= 0.859 Step 8400, Minibatch Loss= 0.7723, Training Accuracy= 0.773 Step 8600, Minibatch Loss= 0.4706, Training Accuracy= 0.859 Step 8800, Minibatch Loss= 0.4800, Training Accuracy= 0.867 Step 9000, Minibatch Loss= 0.4636, Training Accuracy= 0.891 Step 9200, Minibatch Loss= 0.5734, Training Accuracy= 0.828 Step 9400, Minibatch Loss= 0.5548, Training Accuracy= 0.875 Step 9600, Minibatch Loss= 0.3575, Training Accuracy= 0.922 Step 9800, Minibatch Loss= 0.4566, Training Accuracy= 0.844 Step 10000, Minibatch Loss= 0.5125, Training Accuracy= 0.844 Optimization Finished! Testing Accuracy: 0.890625
In [ ]: