NEWS.md

In PhilBoileau/scPCA: Sparse Contrastive Principal Component Analysis

Changes in version 0.99.0 (2019-09-13)

• Submitted to Bioconductor

Changes in version 1.1.2 (2020-01-08)

• Added hierarchical clustering options for clustering based cross-validation

Changes in version 1.1.5 (2020-01-18)

• Fixed issue with matrix normalization
• Misc. bug fixes
• Improvements to code coverage

Changes in version 1.1.11 (2020-02-02)

• Added more SPCA algorithm options
• SPCA via variable projection
• Randomized SPCA via variable projection
• New vignette section comparing performance of SPCA algorithms
• Improvements to code coverage

Changes in version 1.1.12 (2020-04-21)

• Updated citations
• Fixed typos in documentation

Changes in version 1.1.14 (2020-04-26)

• Fixing broken link in an internal function documentation page.

Changes in version 1.1.15 (2020-06-02)

• Fixing Travis CI settings

Changes in version 1.3.2 (2020-08-05)

• Updated scPCA() function documentation
• Corrected spelling mistakes

Changes in version 1.3.3 (2020-08-08)

• The n_centers argument no longer matters when When the contrasts argument is of length 1 and the penalty term is set to 0.
• Users can now pass in their own cluster labels

Changes in version 1.3.4 (2020-08-12)

• Replaced calls to base::eigen() by RSpectra::eigs_sym() to speed up eigendecompositions of contrastive covariance matrices. cPCA is now performed much more quickly when only wishing to compute a handful of leading contrastive principal components.
• Replaced calls to stats::cov() by coop::covar() to speed up computation of large sample covariance matrices.
• In future updates, we'd like to explore using the DelayedArray framework to support the analysis of larger datasets.

Changes in version 1.3.5 (2020-08-18)

• Fixed citations in docs
• Provided more detailed warning when RSpectra::eigs_sym() fails to converge
• Included arguments in scPCA() to control RSpectra::eigs_sym() convergence: error tolerance and max number of iterations

Changes in version 1.3.6 (2020-08-30)

• Fixed issue where n_centers was required when only one penalty and contrast term were provided
• Users can now pass factors and character vectors to the clusters argument.

Changes in version 1.3.7 and 1.3.8 (2020-09-01)

• Minor bug fixes

Changes in version 1.3.9 (2020-10-12)

• scPCA() now accepts DelayedMatrix objects as target and background datasets.

Changes in version 1.3.10 (2020-10-16)

• Implementing suggested improvements from Aaron Lun.

Changes in version 1.5.1 (2020-12-17)

• scPCA() and other internal functions may now take advantage of the ScaledMatrix object class. This allows more computationally efficient contrastive covariance matrix estimation when analyzing large datasets.
• safeColScale() now used MatrixGenerics to handle feature standardization.

Changes in version 1.5.2 (2020-12-21)

• Adding LTLA/ScaledMatrix to "Remotes" section of DESCRIPTION.

Changes in version 1.5.3 (2021-03-14)

• Updating plotting issue in vignette: comparison of cPCA and scPCA loadings.
• Adding pkgdown site.
• Moving ScaledMatrix to "imports" section of DESCRIPTION.

Changes in version 1.7.2 (2021-09-15)

• Removing tests checking that sequential and parallel calls to scPCA() produce identical outputs when BiocParallel's SerialParam() is used. This due to new handing of random number generation in BiocParallel version 1.28.

Changes in version 1.7.3 (2021-10-07)

• Removing more tests attempting to verify that parallelized outputs perfectly match their serial counterparts.

Changes in version 1.9.1 (2022-01-19)

• Updating copyright years
• Updating citation information

Changes in version 1.15.1 (2023-06-19)

• Address useNames issue in colSds() that caused tests to throw warnings.

PhilBoileau/scPCA documentation built on Feb. 6, 2024, 3:33 p.m.