Escaping Parameter Space: Tight Generalization Bounds via Representation Quality

Paper recorded by Signals 4 on 2026-10-16 in cs.LG. Abstract reproduced from arXiv; link to the original below.

Published 2026-10-16 on arXiv · recorded by Signals 4 on 2026-10-19

Category: cs.LG · 机器学习 · first seen 2026-10-19

Abstract

We derive a tight generalization bound for deterministic neural We derive a tight generalization bound for deterministic neural networks based on their last hidden layer representations. By evaluating representation quality with a parameter-free $k$-nearest-neighbor ($k$-NN) classifier, our certificate avoids traditional parameter-space complexity measures, KL divergences to a posterior, weight no

Read on arXiv →