) eval_pts = np.reshape( np.append( eval_pts, (cpp_gp.get_historical_data_copy()).points_sampled[ :, : (cpp_gp_loglikelihood.dim - objective_func._num_fidelity) ], ), ( eval_pts.shape[0] + cpp_gp.num_sampled, cpp_gp.dim - objective_func._num_fidelity, ), ) test = np.zeros(eval_pts.shape[0]) ps_evaluator = PosteriorMean(cpp_gp, num_fidelity) for i, pt in enumerate(eval_pts): ps_evaluator.set_current_point( pt.reshape( (1, cpp_gp_loglikelihood.dim - objective_func._num_fidelity) ) ) test[i] = -ps_evaluator.compute_objective_function() initial_point = eval_pts[np.argmin(test)] ps_sgd_optimizer = cppGradientDescentOptimizer( cpp_inner_search_domain, ps_evaluator, cpp_sgd_params_ps ) report_point = posterior_mean_optimization( ps_sgd_optimizer, initial_guess=initial_point, max_num_threads=4
plt.plot(discrete_pts.flatten(), temp_mean, 'k') plt.fill_between(discrete_pts.flatten(), temp_mean - temp_std, temp_mean + temp_std, color='b', alpha=0.2) x_axis = cpp_gp_nogradient.get_historical_data_copy( ).points_sampled y_axis = cpp_gp_nogradient.get_historical_data_copy( ).points_sampled_value[:, 0] plt.plot(x_axis, y_axis, 'bs') pdf.savefig() # saves the current figure into a pdf page plt.close() ps_evaluator = PosteriorMean(cpp_gp_nogradient, 0) ps_sgd_optimizer = cppGradientDescentOptimizer( cpp_search_domain, ps_evaluator, cpp_sgd_params_ps) fig, ax = plt.subplots(figsize=(figwidth, figheight)) ax.set_ylim(0, 0.5) Y = None #for num, mc in enumerate([20, 100, 1000]): cpp_kg_evaluator_nogradient = cppKnowledgeGradient( gaussian_process=cpp_gp_nogradient, num_fidelity=0, inner_optimizer=ps_sgd_optimizer, discrete_pts=init_points, num_mc_iterations=100) xlist = discrete_pts.flatten()