tamd: Transcendental Algorithm for Mixtures of Distributions
Implements the Transcendental Algorithm for Mixtures
of Distributions (TAMD), a penalized likelihood framework for
fitting finite Gaussian mixture models. TAMD augments the
Expectation-Maximization (EM) algorithm with analytic barrier
terms built from the Hellinger affinity that diverge on the
singular locus, actively preventing component coalescence and
weight degeneracy. Provides the core TAMD fitting function,
closed-form Hellinger affinity and gradient computations, the
Transcendental Affinity Criterion (TAC) for geometry-aware
model selection, the regularity index rho (a scalar diagnostic
for mixture fit quality), and reproduction scripts for all
simulation studies. Methods are described in Fokoue (2024)
<doi:10.48550/arXiv.2602.03889>.
See also Titterington, Smith and Makov (1985,
ISBN:0-471-90510-4) and Watanabe (2009,
ISBN:978-0-521-86408-7).
Documentation:
Downloads:
Linking:
Please use the canonical form
https://CRAN.R-project.org/package=tamd
to link to this page.