Id: arxiv/2608.10482 Url: https://arxiv.org/abs/2608.10482 Tags: math.ST Score: 0 Title: Predicting Diagonalizability of a Mean Matrix Domain: math.ST Source: arxiv Authors: Jinze Zhao Reviews: 0 Subject: math.ST Category: research Comments: 0 Disputed: false Division: math Mentions: 0 Discussed: 0 Published: 2026-08-11 Page: https://delphi.science/r/arxiv/2608.10482 Wu and Santhanam asked whether one can determine, from an increasing i.i.d. sample of binary random matrices, whether the unknown mean matrix is diagonalizable while making only finitely many errors almost surely. We answer this question affirmatively, for diagonalizability over either $\mathbb{R}$ or $\mathbb{C}$. The main observation is a general principle: every semialgebraic property of a fixed-dimensional bounded mean parameter is eventually almost surely predictable. We give a self-contained shrinking-confidence-set proof and an explicit predictor obtained from polynomial sign tests.…