samsam.covis

samsam.covis(mu, cov, logprob, nsamples=100000, print_level=1, print_interval=1000, print_inplace=True, **kwargs)

Importance sampling using a multivariate normal sampling distribution.

The covis sampler generates samples from a normal distribution with mean mu and covariance cov. For each sample, the target distribution logprob is evaluated and a weight is deduced which allows to estimate integrals over the target distribution (e.g. the evidence). The parameters mu and cov should be chosen such that the sampling distribution is close to the target distribution.

Parameters:

mu(ndim,) ndarray

Mean of the sampling distribution.

cov(ndim, ndim) ndarray

Covariacne of the sampling distribution

logprob(x, **kwargs)function

Log probability of the distribution to sample.

nsamples: int

Number of samples to draw.

print_levelint

0 (no printing) 1 (print step)

print_intervalint

Interval at which to print infos.

print_inplacebool

Whether to print infos in place or one line after the other.

**kwargs

Additional parameters for the logprob function.

Returns:

samples(nsamples, ndim) ndarray

Array of parameters values for each sample.

logweights(nsamples,) ndarray

Array of samples log weight.

diagnositicsdict

Dictionary of diagnostics, with the following keys:

logsamp(nsamples,) ndarray

Array of log probability of each sample for the sampling distribution.

logprob(nsamples,) ndarray

Array of log probability of each sample for the target distribution.

logevidencefloat

Log evidence of the target distribution.