Changelog • dineR

dineR 2.0.0

estimation() now supports parallel execution across the regularisation path via the cores argument. Setting cores > 1 distributes each lambda value across worker processes using doSNOW, yielding 2–3× speed-ups on medium-to-large problems (p ≥ 100 or nlambda ≥ 15). Sequential execution (cores = 1) remains the default.
data_generator() now accepts asymmetric sample sizes: n_Y can be specified independently of n_X, allowing the two samples to have different numbers of observations.
All differential network matrices returned by estimation() are now stored as sparse matrices (dgCMatrix class via the Matrix package), reducing memory usage for high-dimensional problems.

Parallelisation — covers how to switch between sequential and parallel modes, documents benchmark results across five problem sizes, and provides guidance on choosing the number of cores.
Estimation — step-by-step walkthrough of data generation and the estimation workflow.
Data Generator — documents the data_generator() function and its outputs in detail.
Differential Networks — end-to-end tutorial on generating data and estimating a differential network.

Fixed summary.estimation() S3 method signature to match the summary generic (object, ...), resolving an R CMD check warning.
Fixed partial argument matching ambiguity in data_generator() where n matched both n_X and n_Y.
Added missing @importFrom foreach foreach %dopar% directive, resolving undefined global variable notes in R CMD check.
Added Matrix and foreach to Imports and doParallel to Suggests in DESCRIPTION.

CRAN release: 2021-11-15