AI Paper Atlas

Research map / Optimization / Nonconvex and smooth optimization

A Class of Parallel Doubly Stochastic Algorithms for Large-Scale Learning

Aryan Mokhtari, Alec Koppel, Martin Takac, Alejandro Ribeiro

JMLR 2020 · Optimization › Nonconvex and smooth optimization · approach: Coordinate and Adam convergence analysis

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
add second-order curvature (ARAPSA instead of RAPSA)RAPSA · Linear regression p=500iterations to near optimum: ARAPSA within 10-4 by t = 300 vs RAPSA within 10 by t = 104Improved
add second-order curvature (ARAPSA instead of RAPSA)RAPSA · MNIST binary N=1e5feature vectors to reach 0.1: tL = 4 × 104 -> fewer than tL = 200Improved
update fewer blocks per iteration (larger B) for RAPSARAPSA · Linear regression p=1024, per feature processedconvergence: B = 128 slightly better than other BImproved
update fewer blocks per iteration (larger B)RAPSA · Linear regression p=1024, per iterationconvergence: best when processors equal blocksHurt
larger B for ARAPSA hybrid step-sizeARAPSA · Linear regression p=1024features to reach 0.1: > 1000 -> 203Improved
larger B for ARAPSA hybrid step-sizeARAPSA · Linear regression p=1024iterations to reach 1e-4: 100 -> > 1000Hurt
larger B for RAPSA hybrid step-sizeRAPSA · MNIST digits 0 vs 8iterations to reach 0.1: 74 -> 631Hurt
larger B for ARAPSA constant step-sizeARAPSA · MNIST digits 0 vs 8iterations to reach 0.1: 145 -> 701Hurt
asynchronous instead of synchronous RAPSARAPSA · Linear regression p=500convergence: synchronized converges fasterHurt
increase clock variability in asynchronous RAPSAAsynchronous RAPSA · MNIST logistic regressionobjective F(xt): no numeric changeNo clear effect
hybrid step-size instead of constant step-size for RAPSARAPSA · MNIST digits 0 vs 8test accuracy: 98% -> 1-2% errorImproved

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