multi_dim=multi_dim)
# Have we already trained a GP that we load or do we train a new one_
if load_params:
gauss.load_model_parameters(train_obs, train_lecs, params_load_path)
else:
train_gp_model(train_obs, train_lecs)
if save_params:
gauss.save_model_parameters(params_save_path, training_tags, kernel, train_parameter_dim,
lengthscale, multi_dim, rescale_data)
# Get prediction from GP
mod_obs, mod_var = calculate_valid(val_lecs)
# Fix plot stuff and stuff
gauss.plot_energy_curve(mod_obs,val_obs, mod_var, val_energy)
gauss.plot_predicted_actual(mod_obs, val_obs, mod_var,
training_tags, validation_tags)
if save_fig:
gauss.generate_and_save_tikz(mod_obs, val_obs, mod_var,
training_tags, validation_tags)
print gauss.get_sigma_intervals(mod_obs, val_obs, mod_var)
gauss.plot_modelerror(val_lecs, train_lecs, mod_obs, val_obs,
training_tags, validation_tags)
print('Model error: ' + str(gauss.get_model_error(mod_obs, val_obs)))
observables = nsopt.get_nsopt_observable(energies, LECM=lecs)
print(
'----------------------------------------------------------------------------------------------------'
)
print(observables)
print(np.size(observables))
print(
'----------------------------------------------------------------------------------------------------'
)
val_lecs = np.split(lecs, 2)[0]
val_energy = np.split(energies, 2)[0]
val_obs = np.split(observables, 2)[0]
train_lecs = np.split(lecs, 2)[1]
train_energy = np.split(energies, 2)[1]
train_obs = np.split(observables, 2)[1]
gauss.set_gp_kernel(kernel=kernel,
in_dim=LEC_LENGTH,
lengthscale=lengthscale,
multi_dim=True)
gauss.populate_gp_model(train_obs,
np.hstack((train_lecs, train_energy)),
rescale=False)
gauss.optimize()
mod_obs, mod_var = gauss.calculate_valid(np.hstack(val_lecs, val_energy))
gauss.plot_predicted_actual(mod_obs, val_obs, mod_var, 'hej', 'hej')