def train_convolutional_neural_network(train, train_labels, test, test_labels,
configuration, shuffle):
tf.reset_default_graph()
x, y, prediction = create_model(configuration)
cost = tf.reduce_mean(
tf.nn.softmax_cross_entropy_with_logits(logits=prediction, labels=y))
optimizer = tf.train.AdamOptimizer(config.LEARNING_RATE).minimize(cost)
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
# Repeat training for N_EPOCHS
for epoch in range(config.N_EPOCHS):
epoch_loss = 0
i = 0
train_np = np.array(train)
train_labels_np = np.array(train_labels)
if (shuffle is True):
# shuffle data in each epoch
train_np, train_labels_np = parse.shuffle_data(
train_np, train_labels_np)
while i < len(train):
start = i
end = i + config.N_BATCH
batch_x = np.array(train_np[start:end], dtype=object)
batch_y = np.array(train_labels_np[start:end])
_, c = sess.run([optimizer, cost],
feed_dict={
x: batch_x,
y: batch_y
})
epoch_loss += c
i += config.N_BATCH
# After each epoch, calculate loss
s = 'Epoch ' + str(epoch) + ', loss: ' + str(epoch_loss)
print(s)
correct = tf.equal(tf.argmax(prediction, 1), tf.argmax(y, 1))
accuracy = tf.reduce_mean(tf.cast(correct, 'float'))
accuracy_value = accuracy.eval({x: test, y: test_labels})
return accuracy_value
def train_neural_network(train, train_labels, test, test_labels, configuration,
shuffle):
tf.reset_default_graph()
# Prepare tensor shapes, hidden layers, weights and biases
x, y, prediction = create_model(configuration)
cost = tf.reduce_mean(
tf.nn.softmax_cross_entropy_with_logits(logits=prediction, labels=y))
optimizer = tf.train.AdamOptimizer(config.LEARNING_RATE).minimize(cost)
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
for epoch in range(config.N_EPOCHS):
epoch_loss = 0
i = 0
# Shuffle data in each epoch
if (shuffle is True):
train, train_labels = parse.shuffle_data(train, train_labels)
# Present whole training set to neural network in batches
while i < len(train):
start = i
end = i + config.N_BATCH
batch_x = np.array(train[start:end])
batch_y = np.array(train_labels[start:end])
_, c = sess.run([optimizer, cost],
feed_dict={
x: batch_x,
y: batch_y
})
# Count lost
epoch_loss += c
i += config.N_BATCH
# After each epoch, calculate loss
s = 'Epoch ' + str(epoch) + ', loss: ' + str(epoch_loss)
print(s)
correct = tf.equal(tf.argmax(prediction, 1), tf.argmax(y, 1))
accuracy = tf.reduce_mean(tf.cast(correct, 'float'))
accuracy_value = accuracy.eval({x: test, y: test_labels})
return accuracy_value