def train(network_architecture, learning_rate= input('Enter the learning rate (default: 0.01 ~ 0.001): '),
#learning_rate = 0,001
batch_size=100, training_epochs=70, display_step=1): #display_step=5):
vae = VariationalAutoencoder(network_architecture,
learning_rate=learning_rate,
batch_size=batch_size)
begin = time.time()
# Training cycle
tempo_total=0.0
for epoch in range(training_epochs):
avg_cost = 0.
total_batch = int(n_samples / batch_size)
for i in range(total_batch):
n = i
batch_xs = importer_.train(batch_size, data1, n)
cost = vae.partial_fit(batch_xs)
#print cost # Compute average loss
avg_cost += cost / n_samples * batch_size
#print avg_cost
# Display logs per epoch step
if epoch % display_step == 0:
end = time.time()
print ("Epoch:", '%04d' % (epoch+1), \
"cost=", "{:.9f}".format(avg_cost),\
" time = %.2fs" % (end - begin))
tempo = (end - begin)
tempo_total = tempo_total + tempo
begin = end
print "tempo total: %.2fs" % tempo_total
reconstructed = vae.reconstruct(batch_xs)
print reconstructed.shape
return vae
n_hidden_recog_2=200, #500
n_hidden_recog_3=30, #27
n_hidden_gener_1=800,
n_hidden_gener_2=200,
n_hidden_gener_3=30,
n_input= num_pix,
n_z=2)
print network_architecture
vae = train(network_architecture, training_epochs = num_epoch, batch_size=num_batch) #75
x_sample = importer_.train(num_batch, data1, 1)
x_reconstruct = vae.reconstruct(x_sample)
def plot_line(num_line):
plt.figure(figsize=(15, 25))
data1 = importer_.plot_linha((x_sample), num_line)
data2 = importer_.plot_linha((x_reconstruct), num_line)
x = np.array(range(len(data1)))
plt.plot( x, data1, 'go') # green bolinha
plt.plot( x, data1, 'k:', color='orange') # linha pontilha orange
plt.plot( x, data2, 'ro') # red triangulo