Package: IBclust 1.5

IBclust: Information Bottleneck Methods for Clustering Mixed-Type Data

Implements multiple variants of the Information Bottleneck ('IB') method for clustering datasets containing continuous, categorical (nominal/ordinal) and mixed-type variables. The package provides deterministic, agglomerative, generalised, sequential, and standard IB clustering algorithms that preserve relevant information while forming interpretable clusters. The Deterministic Information Bottleneck is described in Costa et al. (2026) <doi:10.1016/j.patcog.2026.113580>. The standard IB method originates from Tishby et al. (2000) <doi:10.48550/arXiv.physics/0004057>, the agglomerative variant from Slonim and Tishby (1999) <https://papers.nips.cc/paper/1651-agglomerative-information-bottleneck>, the generalised IB from Strouse and Schwab (2017) <doi:10.1162/NECO_a_00961>, and the sequential IB from Slonim et al. (2002) <doi:10.1145/564376.564401>. Diagnostic and plotting functions are provided to summarise, visualise, and predict from the resulting clusterings.

Authors:Angelos Markos [aut, cre], Efthymios Costa [aut], Ioanna Papatsouma [aut]