AI Paper Atlas

Research map / Optimization / Nonconvex and smooth optimization

Practical Sharpness-Aware Minimization Cannot Converge All the Way to Optima

Dongkuk Si, Chulhee Yun

NeurIPS 2023 · Optimization › Nonconvex and smooth optimization · approach: Coordinate and Adam convergence analysis · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use constant perturbation size rho with gradient normalization instead of decaying rho, for deterministic SAMdeterministic SAM · smooth strongly convex functionsconvergence rate of best-iterate suboptimalityMixed
switch from deterministic to stochastic SAM (constant rho, normalized)SAM · smooth strongly convex functionssuboptimality at convergenceHurt
keep constant rho in deterministic SAM on smooth nonconvex functionsdeterministic SAM · smooth nonconvex functionsaverage squared gradient normHurt
use gradient normalization in the ascent step versus no normalization (ablation)SAM on f(x,y)=(xy-1)^2 · trajectory illustrationlimit pointMixed
use m-SAM (same sample for ascent and descent) instead of n-SAM (independent samples)stochastic SAM · smooth convex functionsconvergence to global minimaHurt
stochastic m-SAM with high gradient noise (sigma > beta rho)m-SAM · smooth strongly convex functionssuboptimality gapHurt
use small gradient noise (sigma <= beta rho) in n-SAM boundstochastic n-SAM · smooth strongly convex and convex functionsconvergence rate modulo additive termImproved
deterministic SAM on nonsmooth Lipschitz convex functions with constant rhodeterministic SAM · nonsmooth Lipschitz convex functionsdistance to global minimumHurt
step size eta of order 1/beta or largerdeterministic SAM · strongly convex quadraticiterate stabilityHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy