Added an optional filematrix row-block provider for benchmark-side
validation artifacts without changing the existing bigmemory PCA backends.
Added true filematrix support for streaming SPCA via
pca_spca_stream_filematrix() and dense block-wise score projection via
pca_scores_stream_filematrix().
Added exact filematrix PCA via pca_stream_filematrix() for moderate-p
benchmark validation. This backend streams row blocks but forms a p x p
covariance matrix, so it is not intended for very-wide p; use
pca_spca_stream_filematrix() for very-wide filematrix workflows.
Added inst/notes/filematrix-backend.md documenting the filematrix backend
strategy, benchmark failure classification, and the decision to defer native
C++ filematrix integration pending R-level benchmarks.
Added vignettes, readme.
Initial CRAN submission.
Renamed the R implementation of scalable PCA to pca_spca_R() and added a
C++-backed pca_spca() along with pca_spca_stream_bigmatrix() for
bigmemory::big.matrix workflows, retaining the block power iteration
algorithm of Elgamal et al. (2015).
Added pca_spca() implementing the scalable PCA algorithm of Elgamal et
al. (2015) with streaming block power iterations for large matrices.
Added an iteratively reweighted singular value decomposition backend for
pca_robust() that exposes the final row weights and iteration count.
Implemented a compiled svd_robust() helper with an svd_robust_R()
reference implementation.