Hands-On Deep Learning Algorithms with Python

Beginning training

Define the batch size and the number of epochs and initialize all the TensorFlow variables:

num_epochs = 100session = tf.InteractiveSession()session.run(tf.global_variables_initializer())

Define a helper function for visualizing results:

def plot(c, x):        c_ = np.argmax(c, 1)    sort_indices = np.argsort(c_, 0)        x_reshape = np.reshape(x[sort_indices], [batch_size, 28, 28])        x_reshape = np.reshape( np.expand_dims(x_reshape, axis=0), [4, (batch_size // 4), 28, 28])    values = []        for i in range(0,4):        row = np.concatenate( [x_reshape[i,j,:,:] for j in range(0,(batch_size // 4))], axis=1)        values.append(row)            return np.concatenate(values, axis=0)