ControlSynth Neural ODEs: Modeling Dynamical Systems with Guaranteed Convergence

TL;DR

ControlSynth Neural ODEs (CSODEs) are a novel class of neural ODEs that guarantee convergence through linear inequalities and incorporate control terms to better model complex physical dynamical systems, especially those described by PDEs.

Contribution

This paper introduces CSODEs, a new neural ODE framework with guaranteed convergence and enhanced modeling capabilities for complex, multi-scale physical systems.

Findings

CSODEs outperform traditional NNs in learning physical dynamics.

Guarantees convergence despite nonlinear properties.

Effective in modeling PDE-formulated systems.

Abstract

Neural ODEs (NODEs) are continuous-time neural networks (NNs) that can process data without the limitation of time intervals. They have advantages in learning and understanding the evolution of complex real dynamics. Many previous works have focused on NODEs in concise forms, while numerous physical systems taking straightforward forms, in fact, belong to their more complex quasi-classes, thus appealing to a class of general NODEs with high scalability and flexibility to model those systems. This, however, may result in intricate nonlinear properties. In this paper, we introduce ControlSynth Neural ODEs (CSODEs). We show that despite their highly nonlinear nature, convergence can be guaranteed via tractable linear inequalities. In the composition of CSODEs, we introduce an extra control term for learning the potential simultaneous capture of dynamics at different scales, which could be particularly useful for partial differential equation-formulated systems. Finally, we compare several representative NNs with CSODEs on important physical dynamics under the inductive biases of CSODEs, and illustrate that CSODEs have better learning and predictive abilities in these settings.