The noise level in linear regression with dependent data (original) (raw)
Abstract:We derive upper bounds for random design linear regression with dependent ($\beta$-mixing) data absent any realizability assumptions. In contrast to the strictly realizable martingale noise regime, no sharp instance-optimal non-asymptotics are available in the literature. Up to constant factors, our analysis correctly recovers the variance term predicted by the Central Limit Theorem -- the noise level of the problem -- and thus exhibits graceful degradation as we introduce misspecification. Past a burn-in, our result is sharp in the moderate deviations regime, and in particular does not inflate the leading order term by mixing time factors.
Submission history
From: Ingvar Ziemann [view email]
[v1] Thu, 18 May 2023 17:55:52 UTC (26 KB)
[v2] Fri, 27 Oct 2023 16:34:44 UTC (62 KB)