+
=
+
=
The futurize package allows you to easily turn sequential code
into parallel code by piping the sequential code to the futurize()
function. Easy!
library(futurize)
plan(multisession)
library(gamlss)
data(abdom, package = "gamlss.data")
cv <- gamlssCV(y ~ pb(x), data = abdom, K.fold = 10) |> futurize()
This vignette demonstrates how to use this approach to parallelize gamlss
functions such as gamlssCV(), add1All(), drop1All(), add1TGD(),
and drop1TGD().
The gamlss package implements Generalized Additive Models for Location, Scale, and Shape (GAMLSS). GAMLSS models extend traditional generalized additive models (GAMs) by allowing all parameters of a distribution — not just the mean — to be modeled as functions of explanatory variables. The package provides tools for model fitting, selection, and diagnostics.
Several gamlss functions support parallel evaluation via the
parallel, ncpus, and cl arguments. By piping to futurize(),
you can leverage any future-based parallel backend for these
computations.
The gamlssCV() function performs k-fold cross-validation for model
selection. This is computationally intensive and benefits greatly
from parallelization:
library(gamlss)
data(abdom, package = "gamlss.data")
cv <- gamlssCV(y ~ pb(x), data = abdom, K.fold = 10)
Here gamlssCV() evaluates each fold sequentially, but we can
easily make it evaluate in parallel by piping to futurize():
library(futurize)
library(gamlss)
data(abdom, package = "gamlss.data")
cv <- gamlssCV(y ~ pb(x), data = abdom, K.fold = 10) |> 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)
Unlike other parallel backends in R, futurize() relays standard
output, messages, and warnings produced by the parallel workers back
to your main R session. For instance, when running the above, you
will see the progress output from each optimizer as it completes:
> cv <- gamlssCV(y ~ pb(x), data = abdom, K.fold = 10) |> futurize()
fold 1
fold 2
fold 3
fold 4
fold 5
fold 6
fold 7
fold 8
fold 9
fold 10
>
This output originates from the parallel workers and is relayed to your R session, so you get the same informative feedback as when running sequentially.
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)
The drop1All() function evaluates the effect of dropping each
term from a fitted GAMLSS model, which can be parallelized:
library(futurize)
plan(multisession)
library(gamlss)
data(abdom, package = "gamlss.data")
m <- gamlss(y ~ pb(x) + x, data = abdom)
d <- drop1All(m) |> futurize()
The following gamlss functions are supported by futurize():
add1All()add1TGD()drop1All()drop1TGD()gamlssCV()For comparison, here is what it takes to parallelize gamlssCV()
using the parallel package directly, without futurize:
library(gamlss)
library(parallel)
data(abdom, package = "gamlss.data")
## Set up a PSOCK cluster
ncpus <- 4L
cl <- makeCluster(ncpus)
clusterEvalQ(cl, library(gamlss))
## Perform k-fold cross-validation in parallel
cv <- gamlssCV(y ~ pb(x), data = abdom, K.fold = 10,
parallel = "snow", ncpus = ncpus, cl = 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().