Parallel linesearch

Surrogate Hessian accelerated parallel line-search: parallel line-search

class stalk.pls.ParallelLineSearch(path='', load=None, *args, **kwargs)

Bases: object

property D

property D_list

property Lambdas

calculate_next_params(N=200, Gs=None, fraction=0.025)

copy(path, structure=None, hessian=None, windows=None, noises=None, pes=None, M=None)

evaluate(add_sigma=False, interactive=False, dep_jobs=[], var_eff_map=None)

evaluate_eqm(add_sigma=False, interactive=False, dep_jobs=[], var_eff_map=None)

property evaluated

property hessian

initialize(windows=None, noises=None, window_frac=None, **ls_args)

ls(i) → LineSearch

property ls_list

ls_type: alias of LineSearch

property noises

property noises_min

property noisy

property params

property params_err

property path

property pes

plot(**kwargs)

propagate(path=None, write=True, overwrite=True, add_sigma=False, fname='pls.p', interactive=False, **kwargs)

property setup

property shifted

property shifts

property structure

property structure_next

property windows

write_to_disk(fname='data.p', overwrite=False)

class stalk.pls.Surrogate(path='', load=None, *args, **kwargs)

Bases: ParallelLineSearch

property Gs

property M

property W_opt

bracket_target_biases(**kwargs)

copy(path='', **kwargs)

property epsilon_d

property epsilon_p

property error_d

property error_p

property ls_list

ls_type: alias of TargetLineSearch

optimize(reoptimize=True, windows=None, noises=None, epsilon_p=None, epsilon_d=None, temperature=None, noise_frac=0.1, resolution=0.01, starting_mix=0.5, write=None, overwrite=False, **ls_args)

optimize_epsilon_d(epsilon_d, Gs=None, skip_setup=False, **ls_args)

optimize_epsilon_p(epsilon_p, starting_mix=0.5, thermal=False, Gs=None, resolution=0.01, verbosity=1, **ls_args)

optimize_temperature(temperature, **ls_args)

optimize_windows_noises(windows, noises, Gs=None, **ls_args)

property optimized

plot_error_surfaces(**kwargs)

property sigma_opt

property statistical_cost

property temperature

validate(N=500, thr=1.1): Validate optimization by independent random resampling

property x_targets