# Training
num_epochs = 100
# Initialize session to make computations on the Tensorflow graph
with tf.Session() as sess:
# Initialize model parameters
sess.run(fis.init_variables)
trn_costs = []
val_costs = []
time_start = time.time()
for epoch in range(num_epochs):
# Run an update step
trn_loss, trn_pred = fis.train(sess, trnData, trnLbls)
# Evaluate on validation set
val_pred, val_loss = fis.infer(sess, chkData, chkLbls)
if epoch % 10 == 0:
print("Train cost after epoch %i: %f" % (epoch, trn_loss))
if epoch == num_epochs - 1:
time_end = time.time()
print("Elapsed time: %f" % (time_end - time_start))
print("Validation loss: %f" % val_loss)
# Plot real vs. predicted
pred = np.vstack((np.expand_dims(trn_pred,
1), np.expand_dims(val_pred, 1)))
plt.figure(1)
plt.plot(mg_series)
plt.plot(pred)
trn_costs.append(trn_loss)
val_costs.append(val_loss)
# Plot the cost over epochs
# Training
num_epochs = 100
# Initialize session to make computations on the Tensorflow graph
with tf.Session() as sess:
# Initialize model parameters
sess.run(fis.init_variables)
trn_costs = []
val_costs = []
time_start = time.time()
for epoch in range(num_epochs):
# Run an update step
trn_loss, trn_pred = fis.train(sess, trnData, trnLbls)
# Evaluate on validation set
val_pred, val_loss = fis.infer(sess, chkData, chkLbls)
if epoch % 10 == 0:
print("Train cost after epoch %i: %f" % (epoch, trn_loss))
if epoch == num_epochs - 1:
time_end = time.time()
print("Elapsed time: %f" % (time_end - time_start))
print("Validation loss: %f" % val_loss)
# Plot real vs. predicted
pred = np.vstack((np.expand_dims(trn_pred, 1), np.expand_dims(val_pred, 1)))
plt.figure(1)
plt.plot(mg_series)
plt.plot(pred)
trn_costs.append(trn_loss)
val_costs.append(val_loss)
# Plot the cost over epochs
plt.figure(2)
# Sets the FIS parameters to the ones on the genome
fis.setmfs(mus, sigmas, y)
pred = fis.infer(data)
loss = 1 - nse(pred, lbls)
return loss
# Runs the evolution cycle
start_time = time.time()
result = list(
differential_evolution(eval_objective,
bounds=[(-2, 2)] * n_params,
gens=10))
end_time = time.time()
print('Evolution time: %f' % (end_time - start_time))
# Gets the last genome
best_params = result[-1][0]
best_mus = best_params[0:fis.m * fis.n]
best_sigmas = best_params[fis.m * fis.n:2 * fis.m * fis.n]
best_y = best_params[2 * fis.m * fis.n:2 * fis.m * fis.n + fis.m]
# Sets the FIS parameters to the ones of the last genome
fis.setmfs(best_mus, best_sigmas, best_y)
# Predicts output for the training set
best_pred = fis.infer(data)
# Plots the real and predicted one series
plt.plot(mg_series)
plt.plot(best_pred)
plt.legend(['Real', 'Predicted'])
plt.show()
print('Best fitness: %f' % result[-1][1])