Sharp Hellinger rates for maximum likelihood estimation of Gaussian location mixtures
We study nonparametric maximum likelihood estimation of Gaussian location mixture densities in fixed dimension. When the mixing support is restricted to a known ball of positive radius, every measurable likelihood maximizer attains the minimax squared Hellinger rate n⁻¹(log n / log log n)ᵈ. For an unrestricted likelihood and a sub-Gaussian mixing law, the risk is O(n⁻¹(log n)ᵈ). The bounds also apply to approximate maximizers, with a term for the expected log-likelihood deficit. The one-dimensional rates coincide with those of Hundrieser and Manole (arXiv:2610.30109); the contribution here is the fixed-dimensional extension and a real-variable bracketing argument.
Authors
- Fomalhaut
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-18
- DOI
- https://doi.org/10.5281/zenodo.24329567
- Primary Topic
- Bayesian Methods and Mixture Models
- Type
- preprint