bayesCL (version 0.0.1)

rpg: Polya-Gamma Random Variates using a GPU

Description

Generate random variates from the Polya-Gamma distribution on a GPU.

Usage

rpg(num=1, n=1, z=0.0, batch=32, local=128, staticseed=FALSE,
seed=0, float=0, ptr=NULL, device=0)

Arguments

num

the number of random variates to simulate.

n

shape parameter, a positive integer.

z

parameter associated with tilting.

batch

the number of samples created by each GPU thread

local

the number of threads in a thread-block on a GPU

staticseed

parameter to determine whether to use a static seed or not.

seed

the value of the static seed, if used.

float

parameter to determine whether to use single-precision floating point or double-precision floating point.

ptr

an external pointer to the C structure with the GPU helper variables.

device

if no external pointer is provided to function, we can provide the ID of the device to use.

Value

This function returns num Polya-Gamma samples.

Details

A random variable X with distribution PG(n,z) is distributed like $$X\sim \sum _{k=1}^{\mathrm{\infty }}G(n,1)/(2{\pi }^2(k-1/2{)}^2+z^2/2).$$ The density for X may be derived by exponentially tilting the PG(n,0) density: $$p(x|n,z)\propto \exp (-x{z}^2/2)p(x|n,0).$$ The GPU implementation is derived from the CPU implementation rpg.devroye from package BayesLogit.

See Also

prepare,mlr

Examples

random_variates <- rpg(num=100, n=1, z=0.0, device=0)