AI Paper Atlas

Research map / Optimization / Optimizers

History-Gradient Aided Batch Size Adaptation for Variance Reduced Algorithms

Kaiyi Ji, Zhe Wang, Bowen Weng, Yi Zhou, Wei Zhang, Yingbin Liang

ICML 2020 · Optimization › Optimizers · approach: Memory-efficient LLM optimizers · 8 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
Adapt epoch batch size using average stochastic gradients from the preceding epochSVRG and SpiderBoost · a8a, a9a, w8a, ijcnn1gradient evaluations to convergenceImproved
Adapt epoch batch size using average stochastic gradients from the preceding epochSVRPG and SPIDER-PG · InvertedPendulum, InvertedDoublePendulum, Swimmer, Hopperconvergence speedImproved
Adapt batch size from history gradients instead of using prescribed exponential or linear growth (ablation)SpiderBoost · a9aloss versus gradient evaluationsImproved
Adapt batch size from history gradients instead of using prescribed growth (ablation)SpiderBoost · w8aloss versus gradient evaluationsMixed
Use exponentially increasing batch size rather than linearly increasing batch size (ablation)SpiderBoost · a9a and w8aloss versus gradient evaluationsMixed
Adapt batch size from history gradientsAbaSVRG · MNIST three-layer neural networktraining lossImproved
Adapt batch size from history gradientsAbaSPIDER · MNIST three-layer neural networkgradient normImproved

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