hamimc-package: A simple Hamiltonian Monte Carlo

In yfyang86/hamimc: Hybrid Monte Carlo

Description

The HMC code is from the original matlab code. Gibbs/MH/ARS methods were added. Besides, a automatically tuning method on stepwise based on original HMC is implemented. Package is under construction, and we use some code that belongs to (Neal,2011). For convience, it is wrappered. This package is not for seriouse research, and has no interests in comparing algorithms.

Details

Package:	hamimc
Type:	Package
Version:	1.1
Date:	2013-11-18
License:	MIT

~~ An overview of how to use the package, including the most important ~~ ~~ functions ~~

Author(s)

Yifan Yang

Maintainer: Who to complain to YANG,Yifan<[e-mail]>

References

Neal, R. "MCMC for Using Hamiltonian Dynamics." Handbook of Markov Chain Monte Carlo (2011): 113-162.

See Also

coda

Examples

## Not run
######################
## M-H tuned
######################
ITER=10000
x.initial=1:50
buinin=5000
dproposalfunction<-function(xt1,xt){# We could try other proposals
  xt<-as.vector(xt)
 Si=diag(rep(1,length(xt)))/20+xt
 mvtnorm::dmvnorm(xt1,mean=xt,sigma=Si,log=T)
}

rproposalfunction<-function(xt){
  xt<-as.vector(xt)
  Si=diag(rep(1,length(xt)))/20+xt
 mvtnorm::rmvnorm(n=1,mean=xt,sigma=Si)
}

#target function
A = matrix(rep(0.998,50^2),ncol=50)*outer(1:50,1:50,'*')
diag(A)=(1:50)^2

dtarget<-function(x,...){
 p=length(x)
 mvtnorm::dmvnorm(x,mean=rep(0,p),sigma=A,log=T)
}

MH.framework(dproposalfunction,rproposalfunction,dtarget,A=A) ->  X.flow

xyplot(mcmc(X.flow[,c(2,45)]))
par(mfrow=c(1,2))
densplot(mcmc(X.flow[-(1:buinin),c(2,45)]))
par(mfrow=c(1,1))

plot(X.flow[-(1:buinin),c(2,45)],type='l')
rejectionRate(mcmc(X.flow[-(1:buinin),c(2,45)]))
#     var1      var2 
# 0.9819964 0.9819964 

## End(**Not run**)

yfyang86/hamimc documentation built on May 4, 2019, 2:32 p.m.