Parallelize 'pls' functions

+ =

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(pls)

data(yarn)
m <- plsr(density ~ NIR, ncomp = 10, data = yarn, validation = "CV") |> futurize()

Introduction

This vignette demonstrates how to use this approach to parallelize pls functions such as mvr(), plsr(), pcr(), and crossval().

The pls package provides Partial Least Squares Regression (PLSR) and Principal Component Regression (PCR) methods. These methods often use cross-validation (CV) to determine the number of components to use, which can be computationally intensive and is an ideal candidate for parallelization.

Example: PLS Regression with Cross-Validation

The plsr() function is used to perform PLS regression. When validation = "CV" is specified, it performs cross-validation.

library(pls)
data(yarn)

## Sequential evaluation
m <- plsr(density ~ NIR, ncomp = 10, data = yarn, validation = "CV")

To make it evaluate in parallel, simply pipe the call to futurize():

library(futurize)
library(pls)
data(yarn)

## Parallel evaluation
m <- plsr(density ~ NIR, ncomp = 10, data = yarn, validation = "CV") |> futurize()

This will automatically use the parallel backend set by plan(), e.g.

plan(multisession)

Example: Stand-alone Cross-Validation

The crossval() function can be used to perform cross-validation on an already fitted model:

library(futurize)
plan(multisession)
library(pls)

data(yarn)
m1 <- plsr(density ~ NIR, ncomp = 10, data = yarn)

## Parallel cross-validation
m_cv <- crossval(m1, segments = 10) |> futurize()

Supported Functions

The following pls functions are supported by futurize():

mvr()
plsr()
pcr()
cppls()
crossval() with seed = TRUE as the default

Without futurize: Manual ‘pls.options’ setup

For comparison, here is what it takes to parallelize pls functions using the parallel package directly, without futurize:

library(pls)
library(parallel)

## Set up a cluster
ncpus <- 4L
cl <- makeCluster(ncpus)

## Configure pls to use the cluster
old_opts <- pls.options(parallel = cl)

## Run regression with cross-validation
data(yarn)
m <- plsr(density ~ NIR, ncomp = 10, data = yarn, validation = "CV")

## Restore original options and stop the cluster
pls.options(old_opts)
stopCluster(cl)