beetroots.approx_optim.approach_type package

Submodules

beetroots.approx_optim.approach_type.abstract_approach_type module

class beetroots.approx_optim.approach_type.abstract_approach_type.ApproachType

Bases: ABC

abstract class to define the optimization method for the likelihood parameter adjustment

abstract optimization()

abstract plot_cost_map()

beetroots.approx_optim.approach_type.bo module

class beetroots.approx_optim.approach_type.bo.BayesianOptimizationApproach

Bases: ApproachType

implements a Bayesian optimization approach to adjust the likelihood parameter

optimization(first_points: List, init_points: int, n_iter: int, list_log10_f_grid: ndarray, pdf_kde_log10_f_Theta: ndarray, pbounds: Dict, sigma_a_val: float, sigma_m_val: float, n: int, ell: int) → None

performs Bayesian optimization

Parameters:

• first_points (List) – list of the points to evaluate before running the optimization

• init_points (int) – number of init points

• n_iter (int) – number of optimization steps to perform

• list_log10_f_grid (np.ndarray) – _description_

• pdf_kde_log10_f_Theta (np.ndarray) – _description_

• pbounds (Dict) – Dictionary with parameters names as keys and a tuple with minimum and maximum values

• sigma_a_val (float) – value of standard deviation of the additive Gaussian noise

• sigma_m_val (float) – value of standard deviation of the multiplicative lognormal noise

• n (int) – pixel index for which the optimization is performed

• ell (int) – line index for which the optimization is performed

plot_GP(bounds_a0_low: ndarray, bounds_a0_high: ndarray, bounds_a1_low: float, bounds_a1_high: float, n_iter: int) → None

plot the state (both mean and standard deviations) of the Gaussian process at multiple steps of the optimization process.

Parameters:

• bounds_a0_low (np.ndarray) – array of lower bounds on the transition location

• bounds_a0_high (np.ndarray) – array of upper bounds on the transition location

• bounds_a1_low (float) – lower bound on the transition slope

• bounds_a1_high (float) – upper bound on the transition slope

• n_iter (int) – number of performed optimization iterations

plot_cost_map()

plot_evaluation_points(bounds_a0_low: ndarray, bounds_a0_high: ndarray, bounds_a1_low: float, bounds_a1_high: float) → None

plots the sequence of sampled points with the order as colorbar

Parameters:

• bounds_a0_low (np.ndarray) – array of lower bounds on the transition location

• bounds_a0_high (np.ndarray) – array of upper bounds on the transition location

• bounds_a1_low (float) – lower bound on the transition slope

• bounds_a1_high (float) – upper bound on the transition slope

plots_postprocessing(bounds_a0_low: ndarray, bounds_a0_high: ndarray, bounds_a1_low: float, bounds_a1_high: float, n_iter: int) → None

makes some plots after the optimization process to facilitate the execution understanding

Parameters:

• bounds_a0_low (np.ndarray) – array of lower bounds on the transition location

• bounds_a0_high (np.ndarray) – array of upper bounds on the transition location

• bounds_a1_low (float) – lower bound on the transition slope

• bounds_a1_high (float) – upper bound on the transition slope

• n_iter (int) – number of performed optimization iterations

beetroots.approx_optim.approach_type.utils module

beetroots.approx_optim.approach_type.utils.compute_lambda(a0: float, a1: float, f_Theta_true: float | ndarray) → float | ndarray

careful: a0 and a1 are in log, base 10, and computations require natural base

Parameters:

• a0 (float) – center of mixing interval

• a1 (float) – radius of mixing interval

• f_Theta_true (Union[float, np.ndarray]) – value of true \(f(\theta)\)

Returns:

lambda – mixing parameter

Return type:

Union[float, np.ndarray]

beetroots.approx_optim.approach_type.utils.estimate_avg_dks_full_bo(a0: float, a1: float, list_log10_f_grid: ndarray, pdf_kde_log10_f_Theta: ndarray, sigma_a: float, sigma_m: float, N_samples_y: int, max_workers: int) → float

beetroots.approx_optim.approach_type.utils.evaluate_cdf(y, a0, a1, f_Theta_true, sigma_a, sigma_m)

beetroots.approx_optim.approach_type.utils.evaluate_pdf(y, a0, a1, f_Theta_true, sigma_a, sigma_m)

beetroots.approx_optim.approach_type.utils.neg_log_pdf_add(y: float | ndarray, m_a: float | ndarray, s_a: float | ndarray) → float | ndarray

evaluates the negative log pdf of a Gaussian distribution

Parameters:

• y (Union[float, np.ndarray]) – points at which the pdf is to be evaluated

• m_a (Union[float, np.ndarray]) – bias of the Gaussian distribution

• s_a (Union[float, np.ndarray]) – standard deviation of the Gaussian distribution

Returns:

negative log pdf

Return type:

Union[float, np.ndarray]

beetroots.approx_optim.approach_type.utils.neg_log_pdf_mult(y, m_m, s_m)

beetroots.approx_optim.approach_type.utils.pdf_my_approx(y, lambda_, m_a, s_a, m_m, s_m, f_Theta_true)

Module contents