log_evidence: Compute the laplace approximation to the log evidence given...

In JimSkinner/spca: Structured PCA

Description

Compute the laplace approximation to the log evidence given the MAP parameters K, mu, sigSq as well as the prior covariance matrix K. Note that this is multiplied by an UN-KNOWN CONSTANT due to the improper priors over mu and sigSq. However, this unknown constant is always the same regardless of k and K, so this may be used to compute meaningful bayes factors between StPCA models.

Usage

log_evidence(X, K, WHat, muHat, sigSqHat, H)

Arguments

X	Data
K	Prior covariance matrix
WHat	Loadings matrix

Value

Approximate log evidence

JimSkinner/spca documentation built on May 7, 2019, 10:52 a.m.