pyselfi.posterior module

Routines related to the SELFI posterior.

class pyselfi.posterior.posterior(theta_0, prior_covariance, inv_prior_covariance, f_0, C_0, inv_C_0, grad_f, phi_obs)[source]

Bases: object

This class represents the SELFI posterior. See equations (25) and (26) in Leclercq et al. (2019) for expressions.

Variables

  • theta_0 (array, double, dimension=S) – prior mean and expansion point

  • prior_covariance (array, double, dimension=(S,S)) – prior covariance matrix

  • inv_prior_covariance (array, double, dimension=(S,S)) – inverse prior covariance matrix

  • f_0 (array, double, dimension=P) – mean blackbox at the expansion point

  • C_0 (array, double, dimension=(P,P)) – covariance matrix of summaries at the expansion point

  • inv_C_0 (array, double, dimension=(P,P)) – inverse covariance matrix of summaries at the expansion point

  • grad_f (array, double, dimension=(S,P)) – gradient of the blackbox at the expansion point

  • phi_obs (array, double, dimension=P) – observed summaries vector

_compute_inv_N()[source]

Computes the inverse of the ‘noise’ matrix.

_compute_inverse_posterior_covariance()[source]

Computes the inverse posterior covariance.

_get_posterior_covariance()[source]

Gets the posterior covariance. See equation (26) in Leclercq et al. (2019).

Returns

Gamma – posterior covariance matrix

Return type

array, double, dimension=(S,S)

_get_posterior_covariance_alt()[source]

Gets the posterior covariance. See equation (26) in Leclercq et al. (2019). Alternative algebra: can be used if numerically more stable.

Returns

Gamma – posterior covariance matrix

Return type

array, double, dimension=(S,S)

_get_posterior_mean()[source]

Gets the posterior mean. See equation (25) in Leclercq et al. (2019).

Returns

gamma – posterior mean

Return type

array, double, dimension=S

_get_posterior_mean_alt()[source]

Gets the posterior mean. See equation (25) in Leclercq et al. (2019). Alternative algebra: can be used if numerically more stable.

Returns

gamma – posterior mean

Return type

array, double, dimension=S

compute()[source]

Computes the posterior (mean and covariance matrix).

classmethod load(fname)[source]

Loads the posterior from an input file.

Parameters

fname (str) – input filename

Returns

posterior – loaded posterior object

Return type

posterior

logpdf(theta, theta_mean, theta_covariance, theta_icov)[source]

Returns the log posterior probability at a given point in parameter space. See equation (24) in Leclercq et al. (2019).

Parameters

  • theta (array, double, dimension=S) – evaluation point in parameter space

  • theta_mean (array, double, dimension=S) – posterior mean

  • theta_covariance (array, double, dimension=(S,S)) – posterior covariance

  • theta_icov (array, double, dimension=(S,S)) – inverse posterior covariance

Returns

logpdf – log posterior probability

Return type

double

save(fname)[source]

Saves the posterior to an output file.

Parameters

fname (str) – output filename