HyperparameterMove {EDISON}

Make a hyperparameter move.

Description

This function makes a hyperparameter move for the information sharing prior selected (or no move if no information sharing prior is selected).

Usage

HyperparameterMove(method, network.info, GLOBvar, hyper.proposals)

Arguments

method

The information sharing method used: 'poisson' for the Poisson prior (no information sharing), 'exp_soft' and 'exp_hard' for the exponential information sharing prior with soft or hard information sharing among nodes, respectively, 'bino_soft' and 'bino_hard' for the binomial information sharing prior with soft or hard information sharing among nodes, respectively.

network.info

Network information collected using CollectNetworkInfo.

GLOBvar

Global variables used during the MCMC.

hyper.proposals

Proposal width for hyperparameters.

Value

List summing up the result of the hypermove. Contains at least:

move.made

Whether a hyperparameter move has been made.

network.info

The network information, possibly updated if the hyperparameter move was made and accepted.

May contain further elements depending on the type of information sharing prior used. See the prior-specific functions ExpHyperMove and BinoHyperMove for details.

Author(s)

Frank Dondelinger

References

For information about the information sharing priors, see:

Husmeier et al. (2010), "Inter-time segment information sharing for non-homogeneous dynamic Bayesian networks", NIPS.

Dondelinger et al. (2012), "Non-homogeneous dynamic Bayesian networks with Bayesian regularization for inferring gene regulatory networks with gradually time-varying structure", Machine Learning.