pacbpred-package: PAC-Bayesian Estimation and Prediction in Sparse Additive...
In pacbpred: PAC-Bayesian Estimation and Prediction in Sparse Additive Models.

Description Details Author(s) References See Also Examples

This package is intended to perform estimation and prediction in high-dimensional additive models, using a PAC-Bayesian point of view and a MCMC algorithm.

Package:	pacbpred
Type:	Package
Version:	0.92.2
Date:	2013-02-05
License:	GPL (>= 2)

Benjamin Guedj

Maintainer: Benjamin Guedj <benjamin.guedj@upmc.fr>

http://www.lsta.upmc.fr/doct/guedj/index.html

Guedj and Alquier (2013), 'PAC-Bayesian Estimation and Prediction in Sparse Additive Models'. Electronic Journal of Statistics, 7, 264–291. DOI:10.1214/13-EJS771. Available on http://projecteuclid.org/DPubS?service=UI&version=1.0&verb=Display&handle=euclid.ejs/1359041592.

pacbpred

ndata <- 100
ntrain <- 80
ntest <- ndata - ntrain
p <- 10
Y <- numeric(ndata)
X <- matrix(nr = ndata, nc = p, data = 2*runif(n = ndata*p) - 1)
for(i in 1:ndata)
  {
    Y[i] <- X[i,1]^3+sin(pi*X[i,2])
  }

Xtrain <- X[1:ntrain,]
Xtest <- X[(ntrain+1):ndata,]
Ytrain <- Y[1:ntrain]
Ytest <- Y[(ntrain+1):ndata]

niter <- 100
cst <- Inf
alpha <- .1
sigma2 <- .1
delta <- ntrain/2

res <- pacbpred(niter = niter, Xtrain = Xtrain, Xtest = Xtest, Y =
Ytrain, cst = cst,
sigma2 = sigma2, delta = delta, alpha = alpha)

print(cbind(res$predict,Ytest))