Benchmarking Energy-Conserving Neural Networks for Learning Dynamics from Data
This work provides a comparative analysis of existing energy-conserving neural networks, which is useful for researchers and practitioners in physics-informed machine learning.
This paper surveys ten energy-conserving neural network models, including HNN, LNN, DeLaN, SymODEN, CHNN, and CLNN, which are designed to learn dynamics from time-series data while enforcing energy conservation. The authors compare their performance across four physical systems and discuss their theoretical underpinnings.
The last few years have witnessed an increased interest in incorporating physics-informed inductive bias in deep learning frameworks. In particular, a growing volume of literature has been exploring ways to enforce energy conservation while using neural networks for learning dynamics from observed time-series data. In this work, we survey ten recently proposed energy-conserving neural network models, including HNN, LNN, DeLaN, SymODEN, CHNN, CLNN and their variants. We provide a compact derivation of the theory behind these models and explain their similarities and differences. Their performance are compared in 4 physical systems. We point out the possibility of leveraging some of these energy-conserving models to design energy-based controllers.