Parallelize 'glmnet' functions

Henrik Bengtsson

+ =

The futurize package allows you to easily turn sequential code into parallel code by piping the sequential code to the futurize() function. Easy!

TL;DR

library(futurize)
plan(multisession)
library(glmnet)

n <- 1000
p <- 100
nzc <- trunc(p / 10)
x <- matrix(rnorm(n * p), n, p)
beta <- rnorm(nzc)
fx <- x[, seq_len(nzc)] %*% beta
eps <- rnorm(n) * 5
y <- drop(fx + eps)

cv <- cv.glmnet(x, y) |> futurize()

Introduction

This vignette demonstrates how to use this approach to parallelize glmnet functions such as cv.glmnet().

The glmnet package uses a highly optimized pathwise coordinate descent algorithm to efficiently compute the entire regularization path for penalized generalized linear models (Lasso, Ridge, Elastic Net). Its cv.glmnet() function performs cross-validation to select the optimal regularization parameter, which is an excellent candidate for parallelization.

Example: Cross-validation for regularized regression

The cv.glmnet() function fits models across multiple folds and lambda values. For example:

library(glmnet)

## Generate simulated data
n <- 1000
p <- 100
nzc <- trunc(p / 10)
x <- matrix(rnorm(n * p), n, p)
beta <- rnorm(nzc)
fx <- x[, seq_len(nzc)] %*% beta
eps <- rnorm(n) * 5
y <- drop(fx + eps)

## Perform cross-validation to find optimal lambda
cv <- cv.glmnet(x, y)

Here cv.glmnet() evaluates sequentially, but we can easily make it evaluate in parallel by piping to futurize():

library(futurize)
library(glmnet)

n <- 1000
p <- 100
nzc <- trunc(p / 10)
x <- matrix(rnorm(n * p), n, p)
beta <- rnorm(nzc)
fx <- x[, seq_len(nzc)] %*% beta
eps <- rnorm(n) * 5
y <- drop(fx + eps)

cv <- cv.glmnet(x, y) |> futurize()

This will distribute the cross-validation folds across the available parallel workers, given that we have set up parallel workers, e.g.

plan(multisession)

The built-in multisession backend parallelizes on your local computer and works on all operating systems. There are other parallel backends to choose from, including alternatives to parallelize locally as well as distributed across remote machines, e.g.

plan(future.mirai::mirai_multisession)

and

plan(future.batchtools::batchtools_slurm)

Supported Functions

The following glmnet functions are supported by futurize():

• cv.glmnet() with seed = TRUE as the default

Without futurize: Manual PSOCK cluster setup

For comparison, here is what it takes to parallelize cv.glmnet() using the parallel and doParallel packages directly, without futurize:

library(glmnet)
library(parallel)
library(doParallel)

## Generate simulated data
n <- 1000
p <- 100
nzc <- trunc(p / 10)
x <- matrix(rnorm(n * p), n, p)
beta <- rnorm(nzc)
fx <- x[, seq_len(nzc)] %*% beta
eps <- rnorm(n) * 5
y <- drop(fx + eps)

## Set up a PSOCK cluster and register it with foreach
ncpus <- 4L
cl <- makeCluster(ncpus)
registerDoParallel(cl)

## Perform cross-validation in parallel via foreach
cv <- cv.glmnet(x, y, parallel = TRUE)

## Tear down the cluster
stopCluster(cl)
registerDoSEQ()  ## reset foreach to sequential

This requires you to manually create a cluster, register it with doParallel, and remember to tear it down and reset the foreach backend when done. If you forget to call stopCluster(), or if your code errors out before reaching it, you leak background R processes. You also have to decide upfront how many CPUs to use and what cluster type to use. Switching to another parallel backend, e.g. a Slurm cluster, would require a completely different setup. With futurize, all of this is handled for you - just pipe to futurize() and control the backend with plan().