English

On the identifiability of Dirichlet mixture models

Statistics Theory 2026-03-24 v1 Statistics Theory

Abstract

We study identifiability of finite mixtures of Dirichlet distributions on the interior of the simplex. We first prove a shift identity showing that every Dirichlet density can be written as a mixture of JJ shifted Dirichlet densities, where J1J-1 is the dimension of the simplex support, which yields non-identifiability on the full parameter space. We then show that identifiability is recovered on a fixed-total parameter slice and on restricted box-type regions. On the full parameter space, we prove that any nontrivial linear relation among Dirichlet kernels must involve at least JJ coefficients sharing a common sign, and deduce that mixtures with fewer than JJ atoms are identifiable. We further report direct non-identifiability implications for unrestricted finite mixtures of generalized Dirichlet, Dirichlet-multinomial, fixed-topic-matrix latent Dirichlet allocation, Beta-Liouville, and inverted Beta-Liouville models.

Cite

@article{arxiv.2603.21914,
  title  = {On the identifiability of Dirichlet mixture models},
  author = {Hien Duy Nguyen and Mayetri Gupta},
  journal= {arXiv preprint arXiv:2603.21914},
  year   = {2026}
}
R2 v1 2026-07-01T11:33:14.201Z