{"id":"arxiv/2609.11273","url":"https://arxiv.org/abs/2609.11273","tags":["math.NA","math.OC","math.ST"],"score":0,"title":"Minimax-Optimal Early Stopping for Continuous-Time SGD via the Discrepancy Principle","domain":"math.NA","source":"arxiv","authors":["Tim Jahn","Loucas Pillaud-Vivien","Adrien Schertzer"],"reviews":0,"subject":"math.NA","summary":"We study early stopping for a continuous-time model of stochastic gradient descent (SGD) in ill-posed linear inverse problems. We consider an a posteriori stopping rule based on the discrepancy principle, which stops the dynamics once the residual reaches the noise level. Unlike deterministic gradient flow, the stochastic dynamics exhibits persistent multiplicative fluctuations whose amplitude scales with the residual itself. Under a source condition on the initial error and a polynomial bound on the spectrum of the empirical covariance operator, we prove that, with high probability, the…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-09-10","reactions":{}}