inst/tinytest/parallel.R

In ncvreg: Regularization Paths for SCAD and MCP Penalized Regression Models

library(parallel, quietly=TRUE)
library(survival, quietly=TRUE)
cl <- makeCluster(4)

# parallel version of cv.ncvreg works

# Linear
X <- matrix(rnorm(500), 50, 10)
y <- rnorm(50)
cvfit <- cv.ncvreg(X, y, cluster=cl)

# Logistic
y <- rbinom(50, 1, 0.5)
cvfit <- cv.ncvreg(X, y, cluster=cl, family='binomial')

# parallel version of cv.ncvsurv works
y <- Surv(rexp(50), sample(rep(0:1, c(10,40))))
X <- matrix(rnorm(50*10), 50, 10)
cvfit <- cv.ncvsurv(X, y, cluster=cl)

# parallel speedup timing
n <- 40
p <- 5000
X <- matrix(rnorm(n*p), ncol=p)
b <- c(2, -2, 1, -1, rep(0, p-4))
y <- rnorm(n, mean=X%*%b, sd=2)
system.time(fit <- cv.ncvreg(X, y, penalty="lasso", nfolds=n, seed=1))
system.time(fit <- cv.ncvreg(X, y, penalty="lasso", nfolds=n, seed=1, cluster=cl))

n <- 40
p <- 5000
X <- matrix(rnorm(n*p), ncol=p)
b <- c(2, -2, 1, -1, rep(0, p-4))
y <- cbind(rexp(n, exp(X%*%b)), rbinom(n, 1, 0.5))
system.time(fit <- cv.ncvsurv(X, y, penalty="lasso", nfolds=n, seed=1))
system.time(fit <- cv.ncvsurv(X, y, penalty="lasso", nfolds=n, seed=1, cluster=cl))

stopCluster(cl)