jexplore.steps.rwalk

jexplore.steps.rwalk#

This module contains definitions of Metropolis-Hasting all chains steps with random walks proposals.

Classes#

`MVRandomWalk`	Generic multivariate random walk proposal.
`GaussianRandomWalk`	Gaussian random walk proposal.
`StudentTRandomWalk`	Student-T random walk proposal.

Module Contents#

class MVRandomWalk(dist, mask=None, scale=1.19, **opts)[source]#

Bases: jexplore.steps.direct.Direct

Generic multivariate random walk proposal.

Parameters:

dist (Type[jexplore.tools.distributions.Distr]) – distribution class
mask (jax.Array | None) – proposal dimensions mask (default all space)
scale (float) – random walk scaling factor.
**opts – options to be passed to the distribution creator.

lower: jax.Array#: covariance cholesky decomposition matrices (nchains, dim, dim).

cd_const: jax.Array#: effective scaling factor. This is the scale attribute divided by the square root of the space dimensionality.

scale: float#: random walk scaling factor.

build(epoch)[source]#

Step epoch initialisation method. This extends method jexplore.steps.step.Step.build by defining the lower and cd_const attributes.

epoch and sampling attributes. :param epoch: current epoch.

Parameters:: epoch (jexplore.sampling.EpochMH)
Return type:: None

proposal(key, state)[source]#

Samples from jexplore.steps.rwalk.MVRandomWalk.epoch_dist are used to perform random walk steps on all chains.

Parameters:

key (jax.Array) – PRNG key used as the random key.
state (jexplore.sampling.StateMH) – a (nwalkers * ntemps, dim) array representing the state point of the MCMC sampler. This corresponds to the jexplore.sampling.state.State.p attribute of a jexplore.sampling.state.State object.

Returns:

the updated PRNG key, the new state point (nwalkers * ntemps, dim) and a (nwalkers *ntemps) array with the log of the ratio between the backward and the forward transition probabilities.

Return type:

tuple[jax.Array, jexplore.sampling.StateMH, jax.Array]

class GaussianRandomWalk(mask=None)[source]#

Bases: MVRandomWalk

Gaussian random walk proposal.

Parameters:: mask (jax.Array | None) – proposal dimensions mask (default all space)

class StudentTRandomWalk(mask=None, nu=5.0)[source]#

Bases: MVRandomWalk

Student-T random walk proposal.

Parameters:

mask (jax.Array | None) – proposal dimensions mask (default all space)
nu (float) – Student-T nu parameter (default: 5)