Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces Cluster-Weighted EDMD (CW-EDMD), an innovative method that enhances the Extended Dynamic Mode Decomposition (EDMD) by learning a soft phase-space partition alongside a tailored EDMD operator for each cluster. By employing an Expectation-Maximization (EM) objective that considers both geometric proximity and prediction residuals, CW-EDMD allows for more accurate local modeling of dynamics in complex systems. The results demonstrate substantial error reductions in predictive performance across various dynamical systems, with median improvements of 57x, 2.7x, and 12x for the damped pendulum, Duffing, and Lorenz systems, respectively.
CW-EDMD achieves up to 57 times better prediction accuracy by tailoring local dynamics modeling to specific state-space regions.
Extended Dynamic Mode Decomposition (EDMD) approximates Koopman operators from data, but a single global operator is inefficient when different state-space regions exhibit distinct local dynamics. We introduce Cluster-Weighted EDMD (CW-EDMD), which jointly learns a soft phase-space partition and a per-cluster EDMD operator. Its Expectation-Maximization (EM) objective assigns each transition based on both geometric proximity and prediction residuals, so clusters specialize where local Koopman models are accurate rather than where the data are dense. On Lorenz, damped pendulum, and Duffing systems, across 36 configurations and 10 seeds, CW-EDMD improves matched-degree EDMD in one-step and 5s-rollout prediction. Across 288 paired comparisons, there are significant error reductions in 258 cases, increases in 4, and no differences in 26. Median one-step error reductions are 57x, 2.7x, and 12x on pendulum, Duffing, and Lorenz, respectively.