{"id":"arxiv/2610.23570","url":"https://arxiv.org/abs/2610.23570","tags":["stat.ML","cs.LG","math.OC","stat.ME"],"score":0,"title":"A general recipe for data-adaptive anytime-valid envelopes for iterative algorithms","domain":"stat.ML","source":"arxiv","authors":["Soham Bonnerjee","Aaditya Ramdas"],"reviews":0,"subject":"stat.ML","summary":"Most convergence guarantees for iterative stochastic algorithms are stated at a fixed iteration count and do not justify stopping at a data-dependent time. Recent work has constructed anytime-valid envelopes that do, but deterministic envelopes are typically too conservative for practical stopping, and existing data-adaptive ones are derived case-by-case, each coupling a specific algorithm to a specific noise model and boundary construction. This paper provides a general recipe for constructing such anytime-valid data-adaptive boundaries for a large class of iterative algorithms. Performance…","category":"research","comments":0,"disputed":false,"division":"stat","mentions":0,"discussed":0,"published":"2026-10-16","reactions":{}}