AI Paper Atlas

Research map / Probabilistic and scientific modeling / Dynamical systems

Lie-Algebraic Acceleration of Neural Koopman Dynamics

Jongwon Lee, Jiwoong Kim, Jungwoo Park, Sungwoo Park

ICML 2026 · Probabilistic and scientific modeling › Dynamical systems · approach: Physics-informed dynamics learning

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 Lie-algebraic parallel-scan Koopman model instead of KoopaNeural Koopman dynamics · Met-Enkephalinbest MSE: 2.17×10−1 -> 7.54×10−2Improved
Heisenberg algebra variant vs unitary variantLie-algebraic Koopman · Met-Enkephalinbest MSE: 7.54×10−2 vs 1.13×10−1Mixed
replace GeoTDM with parallel-scan Lie modelLie-algebraic Koopman · Met-Enkephalininference time (s): 1.38×102 -> 6.36×10−3Improved
replace KoVAE with Lie model on short chainLie-algebraic Koopman · Alkane short chain (C2-C3)best MSE: 5.72×10−2 -> 7.27×10−2Hurt
replace GeoTDM with Lie model on long chainLie-algebraic Koopman · Alkane long chain (C7-C9)best MSE: 5.41×10−2 -> 4.66×10−2Mixed
increase horizon from M=100 to M=1600 (16x) (ablation)Lie-algebraic Koopman · Lotka-Volterrainference time: runtime grows only 6.13×Improved
use parallel scan for long horizons (ablation)Lie-algebraic Koopman · Lotka-Volterramax horizon: up to 1.45 × 106 timestepsImproved

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