Variational Gaussian Approximation in Replica Analysis of Parametric Models

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autor principal: Takahashi, Takashi
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866912707457843200
author Takahashi, Takashi
author_facet Takahashi, Takashi
contents We revisit the replica method for analyzing inference and learning in parametric models, considering situations where the data-generating distribution is unknown or analytically intractable. Instead of assuming idealized distributions to carry out quenched averages analytically, we use a variational Gaussian approximation for the replicated system in grand canonical formalism in which the data average can be deferred and replaced by empirical averages, leading to stationarity conditions that adaptively determine the parameters of the trial Hamiltonian for each dataset. This approach clarifies how fluctuations affect information extraction and connects directly with the results of mathematical statistics or learning theory such as information criteria. As a concrete application, we analyze linear regression and derive learning curves. This includes cases with real-world datasets, where exact replica calculations are not feasible.
format Preprint
id arxiv_https___arxiv_org_abs_2509_11780
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Variational Gaussian Approximation in Replica Analysis of Parametric Models
Takahashi, Takashi
Disordered Systems and Neural Networks
Statistics Theory
Machine Learning
We revisit the replica method for analyzing inference and learning in parametric models, considering situations where the data-generating distribution is unknown or analytically intractable. Instead of assuming idealized distributions to carry out quenched averages analytically, we use a variational Gaussian approximation for the replicated system in grand canonical formalism in which the data average can be deferred and replaced by empirical averages, leading to stationarity conditions that adaptively determine the parameters of the trial Hamiltonian for each dataset. This approach clarifies how fluctuations affect information extraction and connects directly with the results of mathematical statistics or learning theory such as information criteria. As a concrete application, we analyze linear regression and derive learning curves. This includes cases with real-world datasets, where exact replica calculations are not feasible.
title Variational Gaussian Approximation in Replica Analysis of Parametric Models
topic Disordered Systems and Neural Networks
Statistics Theory
Machine Learning
url https://arxiv.org/abs/2509.11780