AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

You Only Sample (Almost) Once: Linear Cost Self-Attention Via Bernoulli Sampling

Zhanpeng Zeng, Yunyang Xiong, Sathya Ravi, Shailesh Acharya, Glenn M Fung, Vikas Singh

ICML 2021 · Architectures and efficiency › Attention and transformers · approach: Linear attention approximations · 3 citations (OpenAlex)

Read the paper

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
Replace softmax attention with YOSO-EBERT-base · MLM pretrainingperplexity: 4.65 -> 4.54Improved
Replace softmax attention with YOSO-EBERT-base · SST-2 devaccuracy: 91.1 -> 92.3Improved
Replace softmax attention with YOSO-EBERT-base · QQP devF1: 87.3 -> 87.3No clear effect
Replace softmax attention with YOSO-EBERT-base · MRPC devF1: 88.3 -> 88.1Hurt
Increase the number of hashes (ablation)YOSO attention · BERT-base GLUE and pretraining validationperformanceImproved
Use the *YOSO backpropagation estimate instead of the standard estimate (ablation)YOSO attention · BERT-base GLUEperformance: 87.3 -> 87.6Improved
Add depthwise convolution (ablation)YOSO attention · LRA averageaccuracy: 58.32 -> 59.16Improved
Replace softmax attention with YOSO-EBERT-small · LRA averageaccuracy: 58.77 -> 59.17Improved
Add depthwise convolution and use the *YOSO gradient estimate (ablation)YOSO attention · LRA averageaccuracy: 58.32 -> 59.87Improved
Replace softmax attention with YOSO attention plus depthwise convolutionBERT-small · PretrainingMLM validation perplexity: 7.05 -> 7.34Hurt
Replace softmax attention with YOSO attention plus depthwise convolutionBERT-small · PretrainingSOP validation accuracy: 91.3% -> 89.6%Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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