GWForge.population_fisher.population_fisher_term_I

The Term-I population Fisher matrix.

Attributes

CONDITION_NUMBER_WARNING
DEFAULT_RCOND
DEGENERACY_THRESHOLD

Classes

PopulationFisherResult

Everything population_fisher() computed, and how to read it.

Functions

covariance_from_fisher(fisher[, names, rcond])	Invert a Fisher matrix, falling back to a pseudo-inverse when it is singular.
population_fisher(model, events[, free_parameters, ...])	Term-I population Fisher matrix for a detected catalogue.

Module Contents

GWForge.population_fisher.population_fisher_term_I.CONDITION_NUMBER_WARNING = 100000000000000.0

GWForge.population_fisher.population_fisher_term_I.DEFAULT_RCOND = 1e-12

GWForge.population_fisher.population_fisher_term_I.DEGENERACY_THRESHOLD = 1e-10

class GWForge.population_fisher.population_fisher_term_I.PopulationFisherResult(parameter_names, fixed_parameters, fiducial, fisher, covariance, condition_number, n_events, n_total, scores, mean_score, degenerate)

Everything population_fisher() computed, and how to read it.

Variables:

• parameter_names (list of str) – Free hyper-parameters, in matrix order.

• fixed_parameters (dict) – Hyper-parameters held fixed, excluded from the matrix.

• fiducial (dict) – Values the Fisher was evaluated at, free and fixed together.

• fisher (numpy.ndarray) – (n, n) Term-I Fisher matrix.

• covariance (numpy.ndarray) – Its inverse (or pseudo-inverse).

• sigma (dict) – Name to one-sigma marginalised uncertainty.

• condition_number (float)

• n_events (int) – Events with finite scores, i.e. those that entered the sum.

• n_total (int) – Events offered, before the support and finiteness cuts.

• scores (numpy.ndarray) – (n_events, n) raw, uncentred scores.

• mean_score (numpy.ndarray) – (n,); the Monte-Carlo estimate of d ln alpha / d Lambda.

• degenerate (list of str) – Free parameters whose centred score vanished.

parameter_names

fixed_parameters

fiducial

fisher

covariance

condition_number

n_events

n_total

scores

mean_score

degenerate

sigma

classmethod from_pickle(path)

Reload a forecast that bin/gwforge_population_fisher saved.

Parameters:

path (str) – A pickle written by the CLI.

Return type:

PopulationFisherResult

correlation_matrix()

Correlation matrix implied by covariance.

scaled_to(n_events)

This forecast rescaled to a different number of detected events.

Parameters:

n_events (int)

Return type:

PopulationFisherResult

summary()

A plain-text table of the forecast.

Return type:

str

GWForge.population_fisher.population_fisher_term_I.covariance_from_fisher(fisher, names=None, rcond=DEFAULT_RCOND)

Invert a Fisher matrix, falling back to a pseudo-inverse when it is singular.

Parameters:

• fisher (numpy.ndarray)

• names (sequence of str or None) – Used only to name parameters in the warning.

• rcond (float)

Returns:

(covariance, condition_number).

Return type:

tuple

GWForge.population_fisher.population_fisher_term_I.population_fisher(model, events, free_parameters=None, fixed_parameters=None, parameters=None, method='auto', rcond=DEFAULT_RCOND, n_total=None)

Term-I population Fisher matrix for a detected catalogue.

Parameters:

• model (GWForge.population_fisher.base.PopulationModel)

• events (dict) – Source parameters of the detected events. They must be drawn from the population at parameters, or the centred-score identity that supplies the selection term does not hold.

• free_parameters (sequence of str or None) – Hyper-parameters to vary. Defaults to every name the model has that is not in fixed_parameters.

• fixed_parameters (dict or None) – Name to value, held fixed and excluded from the matrix.

• parameters (dict or None) – Fiducial point. Defaults to model.fiducial.

• method (str) – Passed to score().

• rcond (float) – Pseudo-inverse threshold.

• n_total (int or None) – Events offered before cuts, for reporting. Defaults to len(events).

Return type:

PopulationFisherResult

Raises:

ValueError – If no parameters are free, or no event has a finite score.