AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Separable Physics-Informed Neural Networks

Junwoo Cho, Seungtae Nam, Hyunmo Yang, Seok-Bae Yun, Youngjoon Hong, Eunbyung Park

NeurIPS 2023 · Architectures and efficiency › RNNs and normalization · approach: Scalable approximation and equivariant compute · 27 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 separated axis-wise networks and forward-mode AD instead of conventional point-wise PINN processingSPINN vs PINN · 3-d PDEs, 64^3 collocation pointsMFLOPs (total): 390,370 -> 280 MFLOPsImproved
Use SPINN instead of causal PINNcausal PINN · (2+1)-d Navier-Stokesruntime (hh:mm): 23:03 -> 00:14Improved
Use SPINN instead of causal PINNcausal PINN · (2+1)-d Navier-Stokesrelative L2 error: 0.0353† -> 0.0355No clear effect
Increase SPINN collocation points from 8^4 to 32^4SPINN · (3+1)-d Navier-Stokesrelative L2 error: 0.0090 -> 0.0019Improved
Increase SPINN rank beyond 128 (ablation)SPINN · (2+1)-d Navier-Stokesrelative L2 errorNo clear effect
Use the modified MLP architecture with SPINN (ablation)SPINN · Diffusion, Helmholtz, and Klein-Gordon PDEsrelative 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