The optimisers.highscore module

class grond.optimisers.highscore.optimiser.DirectedSamplerPhase(*args, **kwargs)[source]

Undocumented.

♦ scatter_scale

float, optional

Scales search radius around the current highscore models

♦ scatter_scale_begin

float, optional

Scaling factor at beginning of the phase.

♦ scatter_scale_end

float, optional

Scaling factor at the end of the directed phase.

♦ starting_point

str (SamplerStartingPointChoice), default: 'excentricity_compensated'

Tunes to the center value of the sampler distribution.May increase the likelihood to draw a highscore member model off-center to the mean value

♦ sampler_distribution

str (SamplerDistributionChoice), default: 'normal'

Distribution new models are drawn from.

♦ standard_deviation_estimator

str (StandardDeviationEstimatorChoice), default: 'median_density_single_chain'

♦ ntries_sample_limit

int, default: 1000

class grond.optimisers.highscore.optimiser.HighScoreOptimiser(**kwargs)[source]

Monte-Carlo-based directed search optimisation with bootstrap.

♦ sampler_phases

list of SamplerPhase objects, default: []

♦ chain_length_factor

float, default: 8.0

♦ nbootstrap

int, default: 100

♦ bootstrap_type

str (BootstrapTypeChoice), default: 'bayesian'

♦ bootstrap_seed

int, default: 23

class grond.optimisers.highscore.optimiser.HighScoreOptimiserConfig(**kwargs)[source]

Undocumented.

♦ sampler_phases

list of SamplerPhase objects, default: [<grond.optimisers.highscore.optimiser.UniformSamplerPhase object at 0x7f24b807ebe0>, <grond.optimisers.highscore.optimiser.DirectedSamplerPhase object at 0x7f24b807ec10>]

Stages of the sampler: Start with uniform sampling of the model model space and narrow down through directed sampling.

♦ chain_length_factor

float, default: 8.0

Controls the length of each chain: chain_length_factor * nparameters + 1

♦ nbootstrap

int, default: 100

Number of bootstrap realisations to be tracked simultaneously in the optimisation.

class grond.optimisers.highscore.optimiser.InjectionSamplerPhase(*args, **kwargs)[source]

Inject predefined/precomputed models into the optimisation

Injected models are given either as an array or are generated from sources/events given in a file. Depending on the problem different sources or events can be used:

Only a single source or event file can be handled in one grond.optimisers.highscore.optimiser.InjectionSamplerPhase. The number of iterations is adjusted according to the number of sources or events found.

♦ xs_inject

numpy.ndarray (pyrocko.guts_array.Array), optional

Array with the injected models.

♦ sources_path

str (grond.meta.Path), optional

File with sources to be injected as models

♦ events_path

str (grond.meta.Path), optional

File with events to be injected as models

class grond.optimisers.highscore.optimiser.SamplerDistributionChoice(...) → dummy for str[source]

Any str out of ['multivariate_normal', 'normal'].

class grond.optimisers.highscore.optimiser.SamplerPhase(*args, **kwargs)[source]

Undocumented.

♦ niterations

int

Number of iteration for this phase.

♦ ntries_preconstrain_limit

int, default: 1000

Tries to find a valid preconstrained sample.

♦ seed

int, optional

Random state seed.

class grond.optimisers.highscore.optimiser.StandardDeviationEstimatorChoice(...) → dummy for str[source]

Any str out of ['median_density_single_chain', 'standard_deviation_all_chains', 'standard_deviation_single_chain'].

class grond.optimisers.highscore.optimiser.UniformSamplerPhase(*args, **kwargs)[source]

Undocumented.