This repository contains the code and figures for our research on analyzing Neural Ordinary Differential Equations (Neural ODEs) from a control-theoretic perspective.
The manuscript has been submitted to the IEEE Conference on Decision and Control (CDC) 2026.
This paper studies Neural ODEs using control-theoretic tools for controllability and observability. A control-affine Neural ODE is analyzed through local linear time-varying Gramians and Koopman lifting. The proposed framework is demonstrated on a series RLC circuit, where the learned model reproduces system trajectories, generalizes to unseen initial conditions, and shows full-rank local controllability and observability Gramians.
.
├── figures/ # Figures related to the research
├── src/ # Source codes to reproduce the results
└── README.md
All figures related to this research are provided in the figures folder. However, only a subset of them is used in the initial manuscript.
All the necessary codes to reproduce our work are available in the src folder.
© 2026 Md Saiful Islam (mi1499@uah.edu) and Rahul Bhadani (rahul.bhadani@uah.edu)