Posterior

posterior

Components for setting up a Log-posterior model from sub-components.

This module provides different components that can be combined into a log-posterior model. Each of them mimics the UM-Bridge interface.

Classes:

Name	Description
GaussianLLFromPTOMap	Gaussian log-likelihood from parameter-to-observable map
LogPosterior	Wrapper for combining log-prior and log-likelihood into a log-posterior

mtmlda.components.posterior.GaussianLLFromPTOMap

Gaussian Log-Likelihood.

This components creates a Gaussian log-likelihood from an UM-Bridge model that constitute a parameter-to-observable (PTO) map.

Methods:

Name	Description
__call__	UM-Bridge-like call interface for the log-likelihood.

__init__

__init__(umbridge_pto_map: ub.Model, data: np.ndarray, covariance: np.ndarray) -> None

Constructor.

Parameters:

Name	Type	Description	Default
umbridge_pto_map	ub.Model	UM-Bridge-server resembling the parameter-to-observable map.	required
data	np.ndarray	Observable data to compute the misfit with, compared to the PTO output	required
covariance	np.ndarray	Covariance for waiting of the misfit vector in Gaussian likelihood	required

Raises:

Type	Description
ValueError	Checks if the sizes of the data vector and the output of the PTO map match
ValueError	Checks if the covariance matrix has the same shape as the data vector

__call__

__call__(parameter: list[list[float]], config: dict[Any]) -> list[list[float]]

UMbridge-like call interface for log-likelihood.

Note that input- and output-formats of this method resemble exactly those in UM-Bridge.

Parameters:

Name	Type	Description	Default
parameter	list[list[float]]	Parameter candidate	required
config	dict[str, Any]	Configuration dictionary to be passed on to UM-Bridge server	required

Returns:

Type	Description
list[list[float]]	list[list[float]]: Log-likelihood value

mtmlda.components.posterior.LogPosterior

__init__

__init__(log_prior: Any, log_likelihood: Any) -> None

Wrapper for computing the log posterior from a log-likelihood and a log-prior component.

log-likelihood and log-prior can be anything, but need to be callable according to the UM-Bridge call interface.

Parameters:

Name	Type	Description	Default
log_prior	Any	Log-prior component	required
log_likelihood	Any	Log-likelihood component	required

__call__

__call__(parameter: list[list[float]], **log_likelihood_args: dict[Any]) -> list[list[float]]

UMbridge-like call interface for log-posterior.

Simply returns the sum of log-likelihood and log-prior values. Note that input- and output-formats of this method resemble exactly those in UM-Bridge. If the log-prior evaluates to -inf, meaning it doesn't have support for the current parameter candidate, likelihood evaluation is not conducted, and -inf returned immediately for the log-posterior.

Parameters:

Name	Type	Description	Default
parameter	list[list[float]]	Parameter candidate	required

Returns:

Type	Description
list[list[float]]	list[list[float]]: Log-posterior value