initFun: Initialization Routines

In thq80/Cornuet_2012_Adaptive-Mutiple-IS: A R Adaptive Multiple Importance Sampling

Description

Functions returning a routine to initialize the IS algorithm. They use a function closure approach in order to accept a general set of arguments and to store in their environment the variables specified by the user. The returned function must be in the form

initialize(N0,p,target,proposal,verbose=FALSE)

Arguments

N0	Sample Size of the Initialization phase
p	Dimension of the sample space.
target	As described in target
proposal	As described in proposal
verbose	To receive updates on the initialization.

Value

Returns a list with components:

w

Vector with initial Importance weights.

xx

Vector with initial Sample.

var

Initial variance.

targ

Initial Value of the target distribution (log-scale)

prop

Initial Value of the proposal distribution (log-scale)

Initialization Procedure

amisInit(maxit=5000,maxVar=100,s=sqrt(maxVar))

Initialization procedure described in Cornuet et al. (2012):

maxit

Maximum number of iterations for the optimization routine;

maxVar

upper bound of the variance.

s

scale parameter.

varInit(Var):

Var

Searches over a range of variances for the one that maximizes the ESS of a sample from the proposal.

uniInit()

Random initialization.

Author(s)

Luca Pozzi, [e-mail]

References

Jean-Marie Cornuet, Jean-Michel Marin, Antonietta Mira and Christian Robert (2012), Adaptive Multiple Importance Sampling, Scandinavian Journal of Statistics

Examples

N0 <- 5
p <- 2
proposal <- mvtComp(df=3)
target <- targetBanana()
amisInit()(N0,p,target,proposal)
varInit(Var=c(10,100))(N0,p,target,proposal)
uniInit()(N0,p,target,proposal)

thq80/Cornuet_2012_Adaptive-Mutiple-IS documentation built on May 21, 2019, 9:23 a.m.