Optimization Guarantees for Square-Root Natural-Gradient Variational Inference

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Kumar, Navish, Möllenhoff, Thomas, Khan, Mohammad Emtiyaz, Lucchi, Aurelien
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866913936605970432
author Kumar, Navish
Möllenhoff, Thomas
Khan, Mohammad Emtiyaz
Lucchi, Aurelien
author_facet Kumar, Navish
Möllenhoff, Thomas
Khan, Mohammad Emtiyaz
Lucchi, Aurelien
contents Variational inference with natural-gradient descent often shows fast convergence in practice, but its theoretical convergence guarantees have been challenging to establish. This is true even for the simplest cases that involve concave log-likelihoods and use a Gaussian approximation. We show that the challenge can be circumvented for such cases using a square-root parameterization for the Gaussian covariance. This approach establishes novel convergence guarantees for natural-gradient variational-Gaussian inference and its continuous-time gradient flow. Our experiments demonstrate the effectiveness of natural gradient methods and highlight their advantages over algorithms that use Euclidean or Wasserstein geometries.
format Preprint
id arxiv_https___arxiv_org_abs_2507_07853
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Optimization Guarantees for Square-Root Natural-Gradient Variational Inference
Kumar, Navish
Möllenhoff, Thomas
Khan, Mohammad Emtiyaz
Lucchi, Aurelien
Machine Learning
Artificial Intelligence
Variational inference with natural-gradient descent often shows fast convergence in practice, but its theoretical convergence guarantees have been challenging to establish. This is true even for the simplest cases that involve concave log-likelihoods and use a Gaussian approximation. We show that the challenge can be circumvented for such cases using a square-root parameterization for the Gaussian covariance. This approach establishes novel convergence guarantees for natural-gradient variational-Gaussian inference and its continuous-time gradient flow. Our experiments demonstrate the effectiveness of natural gradient methods and highlight their advantages over algorithms that use Euclidean or Wasserstein geometries.
title Optimization Guarantees for Square-Root Natural-Gradient Variational Inference
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2507.07853