Changelog • bigPLScox

bigPLScox 0.8.1

Added DOI of the package.
Added high performance PLS Cox backends:
- big_pls_cox_fast() for exact PLS Cox fits on both dense matrices and bigmemory::big.matrix objects.
- big_pls_cox_gd() for gradient based optimisation of Cox partial likelihood in the latent PLS space.
big_pls_cox_gd() now supports several optimisation schemes via the method argument:
- "gd" for a basic fixed step gradient descent,
- "bb" for a Barzilai Borwein step size,
- "nesterov" for Nesterov style acceleration,
- "bfgs" for a quasi Newton update.
All optimisers share the same PLS scores and differ only in how the Cox coefficients are updated.

Fixed problem in C code that led to an additional error during CRAN tests.
Added helpers for big_pls_cox() and big_pls_cox_gd().
New prediction helpers:
- predict.big_pls_cox_fast() and predict.big_pls_cox_gd() now handle dense matrices, big.matrix inputs and in-sample prediction.
- type = "components" returns the PLS scores for the requested components.
- Arguments comps and coef allow partial use of components and user supplied Cox coefficients.
Added simple diagnostic accessors for gradient based fits, including iteration counts, log-likelihood trajectory, gradient norms and step sizes.

See the “Release highlights” section of the README for a condensed overview of these changes.

Added C++ implementations for Cox deviance residuals with streaming support for bigmemory matrices together with benchmarking utilities.
Introduced prediction wrappers and component selection helpers (AIC/BIC) for big_pls_cox() and big_pls_cox_gd().
Enabled naive sparsity control in big_pls_cox() and exposed survival model objects for downstream predictions.
Added cross-validation helpers cv.big_pls_cox() and cv.big_pls_cox_gd() mirroring the plsRcox criteria, including the recommended survivalROC iAUC metric by default.
Documented the legacy and big-memory prediction helpers with runnable examples and cross references to diagnostic utilities.
Extended unit test coverage for the new deviance and prediction features.
Fixed cv.coxgpls() to accept big.matrix predictors without coercion errors.

Added reproducible benchmarking utilities under inst/benchmarks comparing big_pls_cox() against plsRcox::plsRcox() on in-memory and file-backed matrices.
Published two package vignettes covering introductory workflows and large-scale analyses with bigmemory.
Added an introductory vignette covering the core Cox-PLS workflow.
Refreshed the README and website copy to highlight core functionality and to demonstrate working examples without warnings, including guidance on learning materials and benchmarking resources.
Completed package-level documentation with bibliographic references.
Updated package metadata to list optional dependencies used in docs and benchmarks.

Updated maintainer contact details in DESCRIPTION.
Added unit tests for big_pls_cox() and big_pls_cox_gd() stability checks.
Added unit tests covering the new C++-accelerated Cox PLS implementation and cross-validation utilities.

Improved big_pls_cox() numerical stability and added support for additional convergence diagnostics in the gradient-descent solver.
Refactored stochastic gradient solvers to better integrate with bigmemory file-backed matrices.
Improved numerical stability of the deviance residual computations.

Introduced gPLS and sgPLS model families with support for grouped predictors and deviance residual pipelines with cross-validation support.