Parallelize 'rugarch' 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(rugarch)

data(sp500ret, package = "rugarch")
spec <- ugarchspec()
roll <- ugarchroll(spec, sp500ret, n.start = 1000, refit.window = "moving", refit.every = 100) |> futurize()

Introduction

This vignette demonstrates how to use this approach to parallelize rugarch functions such as ugarchroll(), ugarchdistribution(), and ugarchboot().

The rugarch package provides a comprehensive set of methods for Generalized Autoregressive Conditional Heteroskedasticity (GARCH) modeling. Many of its functions, especially those involving rolling estimation or bootstrapping, are computationally intensive and benefit greatly from parallelization.

Example: Rolling GARCH estimation

The ugarchroll() function performs rolling estimation and forecasting. This can be time-consuming as it involves multiple fits of the GARCH model.

library(futurize)
plan(multisession)
library(rugarch)

data(sp500ret, package = "rugarch")
spec <- ugarchspec()

## Perform rolling estimation
roll <- ugarchroll(spec, sp500ret, n.start = 1000, 
                   refit.window = "moving", refit.every = 100) |> futurize()

Example: GARCH parameter distribution

The ugarchdistribution() function simulates and estimates the parameter distribution of a GARCH model.

library(futurize)
plan(multisession)
library(rugarch)

data(sp500ret, package = "rugarch")
spec <- ugarchspec()
fit <- ugarchfit(spec, sp500ret)

## Estimate parameter distribution
dist <- ugarchdistribution(fit, n.sim = 100, n.hist = 10) |> futurize()

Supported Functions

The following rugarch functions are supported by futurize():

arfimacv() with seed = TRUE as the default
arfimadistribution() with seed = TRUE as the default
arfimaroll() with seed = TRUE as the default
autoarfima() with seed = TRUE as the default
multifilter() with seed = TRUE as the default
multifit() with seed = TRUE as the default
multiforecast() with seed = TRUE as the default
ugarchboot() with seed = TRUE as the default
ugarchdistribution() with seed = TRUE as the default
ugarchroll() with seed = TRUE as the default

Without futurize: Manual PSOCK cluster setup

For comparison, here is what it takes to parallelize ugarchroll() using the parallel package directly, without futurize:

library(rugarch)
library(parallel)

data(sp500ret, package = "rugarch")
spec <- ugarchspec()

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

## Run rolling estimation in parallel
roll <- ugarchroll(spec, sp500ret, n.start = 1000, 
                   refit.window = "moving", refit.every = 100,
                   cluster = cl)

## Tear down the cluster
stopCluster(cl)

This requires you to manually create and manage the cluster lifecycle. 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().