scaling=False,
verbose=True)
# Run the Bayessian optimization
Bopt = bo_optimizer.optimize(test_function, n_iter=nb_iter_bo)
print(Bopt)
if display_figures:
# Plot the acquisition function
fig = plt.figure(figsize=(10, 10))
ax = Axes3D(fig)
bo_plot_acquisition_spd(
ax,
acq_fct,
r_cone=r_cone,
xs=bo_optimizer.acquisition.data[0],
opt_x=Bopt.x,
true_opt_x=symmetric_matrix_to_vector_mandel(true_sigma)[None],
alpha=0.25,
n_elems=20,
n_elems_h=10)
ax.set_title('Acquisition function', fontsize=50)
plt.show()
# Plot the GP
fig = plt.figure(figsize=(10, 10))
ax = Axes3D(fig)
bo_plot_gp_spd(
ax,
model,
r_cone=r_cone,
xs=bo_optimizer.acquisition.data[0],
# BO optimum to SPD
opt_x_chol = np.zeros((dim, dim))
opt_x_chol[indices] = Bopt.x
opt_x_mat = np.dot(opt_x_chol, opt_x_chol.T)
opt_x_vec = symmetric_matrix_to_vector_mandel(opt_x_mat)[None]
if display_figures:
# Plot the acquisition function
fig = plt.figure(figsize=(10, 10))
ax = Axes3D(fig)
bo_plot_acquisition_spd(
ax,
acq_fct,
r_cone=r_cone,
xs=x_eval_vec,
opt_x=opt_x_vec,
true_opt_x=symmetric_matrix_to_vector_mandel(true_sigma)[None],
chol=True,
alpha=0.25,
n_elems=20,
n_elems_h=10)
ax.set_title('Acquisition function', fontsize=50)
plt.show()
# Plot the GP
fig = plt.figure(figsize=(10, 10))
ax = Axes3D(fig)
bo_plot_gp_spd(
ax,
model,
r_cone=r_cone,
x_eval_mat[n] = np.dot(sigmaL, sigmaL.T)
x_eval[n] = symmetric_matrix_to_vector_mandel(x_eval_mat[n])
best_x_mat = np.zeros((len(best_x_chol), dim, dim))
best_x = np.zeros((len(best_x_chol), dim_vec))
for n in range(len(best_x_chol)):
sigmaL = np.zeros((dim, dim))
sigmaL[indices] = best_x_chol[n].detach().numpy()
best_x_mat[n] = np.dot(sigmaL, sigmaL.T)
best_x[n] = symmetric_matrix_to_vector_mandel(best_x_mat[n])
if disp_fig:
# Plot acquisition function
fig = plt.figure(figsize=(5, 5))
ax = Axes3D(fig)
bo_plot_acquisition_spd(ax, acq_fct, r_cone=r_cone, xs=x_eval, opt_x=best_x[-1][None],
true_opt_x=true_min_vec, n_elems=20, n_elems_h=10)
ax.set_title('Acquisition function', fontsize=20)
plt.show()
# Plot GP
fig = plt.figure(figsize=(5, 5))
ax = Axes3D(fig)
bo_plot_gp_spd(ax, model, r_cone=r_cone, xs=x_eval, opt_x=best_x[-1][None], true_opt_x=true_min_vec,
true_opt_y=true_opt_val, max_colors=25., n_elems=20,
n_elems_h=10)
ax.set_title('GP mean', fontsize=20)
plt.show()
if dim == 2:
# Plot convergence on the sphere
# 3D figure