parallel: Get initial values for parallel RNGs

In rjags: Bayesian Graphical Models using MCMC

Get initial values for parallel RNGs

Description

On a multi-processor system, you may wish to run parallel chains using multiple jags.model objects, each running a single chain on a separate processor. This function returns a list of values that may be used to initialize the random number generator of each chain.

Usage

parallel.seeds(factory, nchain)

Arguments

factory	Name of the RNG factory to use.
nchain	Number of chains for which to initialize RNGs.

Value

parallel.seeds returns a list of RNG states. Each element is a list of length 2 with the following elements:

.RNG.name	The name of the RNG
.RNG.state	An integer vector giving the state of the RNG.

Note

It is not yet possible to make the results of parallel.seeds reproducible. This will be fixed in a future version of JAGS.

Author(s)

Martyn Plummer

See Also

jags.model, section “Random number generators”, for further details on RNG initialization; list.factories to find the names of available RNG factories.

Examples

##The BaseRNG factory generates up to four distinct types of RNG. If
##more than 4 chains are requested, it will recycle the RNG types, but
##use different initial values
parallel.seeds("base::BaseRNG", 3)

## The lecuyer module provides the RngStream factory, which allows large
## numbers of independent parallel RNGs to be generated. 
load.module("lecuyer")
list.factories(type="rng")
parallel.seeds("lecuyer::RngStream", 5);

rjags documentation built on Sept. 11, 2024, 6:31 p.m.