AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

O(n) Connections are Expressive Enough: Universal Approximability of Sparse Transformers

Chulhee Yun, Yin-Wen Chang, Srinadh Bhojanapalli, Ankit Singh Rawat, Sashank Reddi, Sanjiv Kumar

NeurIPS 2020 · Architectures and efficiency › Attention and transformers · approach: Sparse efficient transformers for long context · 10 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 a deterministic sparsity pattern instead of RandomSparse Transformers · Synthetic copying taskaccuracyImproved
Use Star attention rather than Strided or Fixed attentionSparse Transformers · One Billion Word BenchmarkperplexityImproved
Use Star attention rather than the other tested patternsSparse Transformers · Synthetic copying taskaccuracyHurt
Use Union rather than Sequential or Multihead configurationStrided and Fixed sparse Transformers · WMT18 English-Czech, Newstest 2015BLEUImproved
Increase sparsity in the Multihead configurationStrided and Fixed sparse Transformers · WMT18 English-Czech, Newstest 2015BLEUImproved
Use Union rather than the other head configurationsStrided and Fixed sparse Transformers · BERT tasksaccuracyHurt
Choose sparsity pattern based on task rather than assuming one pattern is best everywhereSparse Transformers · NLP taskstask performanceMixed

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