{"id":"arxiv/2609.32962","url":"https://arxiv.org/abs/2609.32962","tags":["cs.LG","math.OC"],"score":0,"title":"Clipped or Unclipped? Finite-Sample Trade-offs for Averaged SGD under Heavy-Tailed Noise","domain":"cs.LG","source":"arxiv","authors":["Alexandra Suvorikova","Egor Gladin","Darina Dvinskikh","Artem Agafonov","Mohammad Alkousa","Yuriy Dorn","Vladislav Matyukhin","Alexander Gasnikov"],"reviews":0,"subject":"cs.LG","summary":"Gradient clipping is widely used to stabilize training, but it need not improve the statistical accuracy of averaged SGD, even under heavy-tailed noise. We derive a finite-sample comparison of clipped and unclipped Polyak-Ruppert averaged SGD under finite conditional $p$-th moments, $p\\ge2$. Our main result gives explicit accuracy and confidence conditions under which, for $p>2$, the Gaussian term dominates the unclipped deviation bound, so clipping need not improve its leading order. By balancing clipping bias and concentration, we obtain a bound in which the heavy-tail correction depends…","category":"research","comments":0,"disputed":false,"division":"cs","mentions":0,"discussed":0,"published":"2026-09-26","reactions":{}}