AI Paper Atlas

Research map / Optimization / Optimizers

Memory Efficient Adaptive Optimization

Rohan Anil, Vineet Gupta, Tomer Koren, Yoram Singer

NeurIPS 2019 · Optimization › Optimizers · approach: Memory-efficient LLM optimizers · 18 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 SM3 instead of AdamTransformer-Big · WMT'14 en→fr, batch 384BLEU: 38.96 ± 0.002 -> 39.81 ± 0.002Improved
Use SM3 instead of AdafactorTransformer-Big · WMT'14 en→fr, batch 768BLEU: 39.65 ± 0.002 -> 40.50 ± 0.001Improved
Increase batch size from 384 to 768 using SM3SM3 on Transformer-Big · WMT'14 en→frBLEU: 39.81 ± 0.002 -> 40.50 ± 0.001Improved
Use SM3 instead of AdamBERT-Large · Wikipedia+BooksCorpusMasked-LM test accuracyNo clear effect
Increase batch size from 1024 to 2048 using SM3SM3 on BERT-Large · Wikipedia+BooksCorpusMasked-LM test accuracyImproved
Use a double batch size with SM3BERT-Large · Wikipedia+BooksCorpus, target 70% Masked-LM accuracywall-time to target: almost 35% less wall-timeImproved
Use SM3 instead of SGD with momentumAmoebaNet-D · ImageNettop-1 accuracy: 78.71%Improved
Use SM3-II rather than SM3-I (ablation)SM3 · Adagrad accumulator approximationapproximation tightnessImproved

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