AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Learning Operators with Coupled Attention

Georgios Kissas, Jacob Seidman, Leonardo Ferreira Guilhoto, Victor M. Preciado, George J. Pappas, Paris Perdikaris

NeurIPS 2022 · Architectures and efficiency › Attention and transformers · approach: Theory of self-attention dynamics

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 kernel-coupled attention (KCA) weights (ablation)LOCA · Darcy flow, 1.5% labeled, N=200mean relative L2 error: 0.463 -> 0.017Improved
add KCA weights (ablation)LOCA · Darcy flow, 1.5% labeled, N=200max relative L2 error: 3.023 -> 0.041Improved
LOCA vs DON and FNO with 6% labeled outputsLOCA · Darcy flowrelative L2 errorImproved
LOCA vs DON and FNO at 2.5% labeled dataLOCA · Shallow Waterrelative L2 errorMixed
clean train, noisy test: LOCA vs FNOLOCA · Mechanical MNISTpercent error increase v1: 1,930% -> 80%Improved
clean train, noisy test: LOCA vs DeepONetLOCA · Mechanical MNISTpercent error increase v1: 112% -> 80%Improved
noisy train and test: LOCA vs FNOLOCA · Mechanical MNISTpercent error increase v1: 280% -> 26%Improved
noisy train and test: LOCA vs DeepONetLOCA · Mechanical MNISTpercent error increase v1: 128% -> 26%Improved
clean train and clean test: LOCA vs FNOLOCA · Mechanical MNISTmean relative L2 error v1: 0.004 -> 0.026Hurt
clean train and clean test: LOCA vs DeepONetLOCA · Mechanical MNISTmean relative L2 error v1: 0.028 -> 0.026No clear effect
different random initializations (10 seeds)LOCA · Mechanical MNISTmax relative L2 errorMixed
LOCA vs DON and FNO in temporal extrapolationLOCA · Climate modeling, 4% labeledrelative L2 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