New SADFED framework identifies Koopman models without predefined architectures

Researchers have developed a new framework called Spatially Aware Dictionary-Free Koopman Eigenfunction Identification (SADFED) for discovering Koopman models from data. This method does not require pre-defining a lifting dictionary, kernel, or neural network architecture. SADFED uses regularized least squares to identify Koopman modes and then employs a transformed temporal basis to obtain eigenfunction values, optimizing only the real and imaginary parts of eigenvalues. The framework was tested on various systems, including the FitzHugh-Nagumo system, the van der Pol oscillator, the Duffing system, and a two-spool turbojet engine, demonstrating its ability to recover known eigenfunctions and construct state-dependent lifted input dynamics for control applications. AI

IMPACT This research introduces a novel method for modeling complex dynamical systems, potentially improving control and simulation capabilities in AI applications.

RANK_REASON The cluster contains a research paper detailing a new method for identifying Koopman models. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• arXiv
• David Grasev
• Duffing system
• FitzHugh-Nagumo system
• Koopman
• Spatially Aware Dictionary-Free Koopman Eigenfunction Identification
• two-spool turbojet engine
• Van der Pol oscillator

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →