gipsDA: Training DA Models Utilizing 'gips'

Extends classical linear and quadratic discriminant analysis by incorporating permutation group symmetries into covariance matrix estimation. The package leverages methodology from the 'gips' framework to identify and impose permutation structures that act as a form of regularization, improving stability and interpretability in settings with symmetric or exchangeable features. Several discriminant analysis variants are provided, including pooled and class-specific covariance models, as well as multi-class extensions with shared or independent symmetry structures. For more details about 'gips' methodology see and Graczyk et al. (2022) <doi:10.1214/22-AOS2174> and Chojecki, Morgen, Kołodziejek (2025, <doi:10.18637/jss.v112.i07>).

install.packages("gipsDA")

Try the gipsDA package in your browser