AI Paper Atlas

Research map / Optimization / Nonconvex and smooth optimization

Stochastic Conditional Gradient Methods: From Convex Minimization to Submodular Maximization

Aryan Mokhtari, Hamed Hassani, Amin Karbasi

JMLR 2020 · Optimization › Nonconvex and smooth optimization · approach: Stochastic gradient last-iterate analysis · 65 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
average stochastic gradients over time (SFW) instead of using a single mini-batch gradientMini-batch FW · Stochastic quadratic program (n = 5, sigma = 100)suboptimality gapImproved
use SFW with b = 1 instead of mini-batch FW with b = 100Mini-batch FW · Stochastic quadratic program (sigma = 300)suboptimality gapImproved
use SFW with b = 10 instead of mini-batch FW with b = 1000Mini-batch FW · Matrix completion (n = 200)normalized error: 0.55 -> 0.25Improved
increase SFW batch size from 10 to 1000 (ablation)SFW · Matrix completion (n = 200)normalized error: 0.25 -> 2.3\times 10^{-3}Improved
use growing batch size O(t^2) instead of SFWSFW · Matrix completion (n = 200), per iterationnormalized errorHurt
use SFW instead of growing-batch FW when comparing by samples processedGrowing-batch FW · Matrix completion (n = 200), per samplenormalized errorImproved
set averaging parameter rho_t = O(t^{-2/3}) instead of t^{-1/3}, t^{-1/2} or t^{-1} (ablation)SFW · Matrix completion (n = 200)convergenceImproved

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