DissipNet: Data-driven Discrete-time Deep RNN-based Modeling for Dissipative Systems

This paper introduces DissipNet, a deep discrete-time dissipative recurrent neural network designed for modeling dissipative dynamical systems. The model explicitly enforces dissipativity, a fundamental physical property related to system stability and energy dissipation. By incorporating physical constraints, DissipNet aims to achieve high prediction accuracy while preserving essential characteristics of real-world dynamical systems in applications such as control theory and complex dynamics analysis.