Spherical Priors, Repeated Gaussian Products, and Bayes Risk

We study matrix-normal mean estimation with spherical priors, local density limits for real Gaussian matrix products, and covariance inference from partial observations. An exact eigenvalue criterion characterizes minimaxity under row-weighted quadratic loss. For products of fixed depth with repeated critical intermediate widths, we derive local density and small-ball asymptotics and transfer them to posterior distributions and Bayes risk. We characterize which covariance components can be identified from products of spherical averages and derive Fisher information rates near rank loss. For paired orthogonal readouts of the same three-by-three matrix, we determine the optimal efficient information for determinant-linear perturbations with the other quartic parameters unknown. A finite set of rotations attains the Haar design information at the Gaussian reference law.Version 2 adds a guide to the results and a notation table, clarifies the scope, and improves copying of matrix-order symbols in the PDF. Using the sign symmetry of the paired readings, the revised construction preserves the full quartic Fisher information at the Gaussian baseline with at most 18,225 ordered frame pairs. The proof and computation supplement give the reduced rotation rule.

Authors

Publication Details

Journal
Figshare
Published
2026-11-20
DOI
https://doi.org/10.6084/m9.figshare.35076490
Primary Topic
Random Matrices and Applications
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Spherical Priors, Repeated Gaussian Products, and Bayes Risk

Akihiro Koide
Figshare
Random Matrices and Applications
preprint

Spherical Priors, Repeated Gaussian Products, and Bayes Risk

Akihiro Koide
preprint en

Abstract

We study matrix-normal mean estimation with spherical priors, local density limits for real Gaussian matrix products, and covariance inference from partial observations. An exact eigenvalue criterion characterizes minimaxity under row-weighted quadratic loss. For products of fixed depth with repeated critical intermediate widths, we derive local density and small-ball asymptotics and transfer them to posterior distributions and Bayes risk. We characterize which covariance components can be identified from products of spherical averages and derive Fisher information rates near rank loss. For paired orthogonal readouts of the same three-by-three matrix, we determine the optimal efficient information for determinant-linear perturbations with the other quartic parameters unknown. A finite set of rotations attains the Haar design information at the Gaussian reference law.Version 2 adds a guide to the results and a notation table, clarifies the scope, and improves copying of matrix-order symbols in the PDF. Using the sign symmetry of the paired readings, the revised construction preserves the full quartic Fisher information at the Gaussian baseline with at most 18,225 ordered frame pairs. The proof and computation supplement give the reduced rotation rule.

Figshare
Random Matrices and Applications
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon