plt.xlabel('z1 (mean)')
plt.ylabel('z2 (mean)')
plt.show()
# plot the mode of responsibilities
plt.scatter(modes[:, 0], modes[:, 1], c=variables_train[:, y_number])
plt.colorbar()
plt.ylim(-1.1, 1.1)
plt.xlim(-1.1, 1.1)
plt.xlabel('z1 (mode)')
plt.ylabel('z2 (mode)')
plt.show()
# GTMR prediction
# Forward analysis (Regression)
mean_of_estimated_mean_of_y, mode_of_estimated_mean_of_y, responsibilities_y, py = \
model.gtmr_predict(autoscaled_variables_test[:, numbers_of_x], numbers_of_x, numbers_of_y)
# Inverse analysis
mean_of_estimated_mean_of_x, mode_of_estimated_mean_of_x, responsibilities_y, py = \
model.gtmr_predict(autoscaled_variables_test[:, numbers_of_y], numbers_of_y, numbers_of_x)
# Check results of forward analysis (regression)
print('Results of forward analysis (regression)')
predicted_y_test_all = mode_of_estimated_mean_of_y.copy()
plt.rcParams['font.size'] = 18
for y_number in range(len(numbers_of_y)):
predicted_y_test = np.ndarray.flatten(predicted_y_test_all[:,
y_number])
predicted_y_test = predicted_y_test * variables_train[:, numbers_of_y[
y_number]].std(
ddof=1) + variables_train[:, numbers_of_y[y_number]].mean()
autoscaled_target_y_value = (target_y_value - variables.mean(
axis=0)[numbers_of_y]) / variables.std(axis=0, ddof=1)[numbers_of_y]
# construct GTMR model
model = GTM(shape_of_map, shape_of_rbf_centers, variance_of_rbfs,
lambda_in_em_algorithm, number_of_iterations, display_flag)
model.fit(autoscaled_variables)
if model.success_flag:
# calculate of responsibilities
responsibilities = model.responsibility(autoscaled_variables)
means = responsibilities.dot(model.map_grids)
modes = model.map_grids[responsibilities.argmax(axis=1), :]
mean_of_estimated_mean_of_y, mode_of_estimated_mean_of_y, responsibilities_y, py = \
model.gtmr_predict(autoscaled_variables.iloc[:, numbers_of_x], numbers_of_x, numbers_of_y)
plt.rcParams['font.size'] = 18
for index, y_number in enumerate(numbers_of_y):
predicted_y_test = mode_of_estimated_mean_of_y[:, index] * variables.iloc[:, y_number].std(
) + variables.iloc[:, y_number].mean()
# yy-plot
plt.figure(figsize=figure.figaspect(1))
plt.scatter(variables.iloc[:, y_number], predicted_y_test)
y_max = np.max(
np.array([np.array(variables.iloc[:, y_number]),
predicted_y_test]))
y_min = np.min(
np.array([np.array(variables.iloc[:, y_number]),
predicted_y_test]))
plt.plot(
plt.xlim(-1.1, 1.1)
plt.xlabel('z1 (mean)')
plt.ylabel('z2 (mean)')
plt.show()
# plot the mode of responsibilities
plt.scatter(modes[:, 0], modes[:, 1], c=variables[:, y_number])
plt.colorbar()
plt.ylim(-1.1, 1.1)
plt.xlim(-1.1, 1.1)
plt.xlabel('z1 (mode)')
plt.ylabel('z2 (mode)')
plt.show()
# GTMR prediction for inverse analysis
mean_of_estimated_mean_of_x, mode_of_estimated_mean_of_x, responsibilities_y, py = \
model.gtmr_predict(autoscaled_target_y_value, numbers_of_y, numbers_of_x)
# Check results of inverse analysis
print('Results of inverse analysis')
mean_of_estimated_mean_of_x = mean_of_estimated_mean_of_x * x.std(
axis=0, ddof=1) + x.mean(axis=0)
mode_of_estimated_mean_of_x = mode_of_estimated_mean_of_x * x.std(
axis=0, ddof=1) + x.mean(axis=0)
# print('estimated x-mean: {0}'.format(mean_of_estimated_mean_of_x))
print('estimated x-mode: {0}'.format(mode_of_estimated_mean_of_x))
estimated_x_mean_on_map = responsibilities_y.dot(model.map_grids)
estimated_x_mode_on_map = model.map_grids[np.argmax(responsibilities_y), :]
# print('estimated x-mean on map: {0}'.format(estimated_x_mean_on_map))
print('estimated x-mode on map: {0}'.format(estimated_x_mode_on_map))