JSINDy (Joint SINDy) is an approach to jointly learning sparse governing equations and system states from limited, incomplete, and noisy observations. JSINDy combines sparse recovery strategies over a function library, a least squares collocation approach to solving ODEs, and reproducing kernel Hilbert space (RKHS) regularization to simultaneously fit dynamics and estimate trajectories.
Paper: A joint optimization approach to identifying sparse dynamics using least squares kernel collocation
Requires Python >= 3.10, and you may want to manually install jax[cuda12] first before pip install -e .
A. W. Hsu, I. Griss Salas, J. M. Stevens-Haas, J. N. Kutz, A. Aravkin, and B. Hosseini, "A joint optimization approach to identifying sparse dynamics using least squares kernel collocation," arXiv preprint arXiv:2511.18555, 2025.
@misc{hsu2025jointoptimizationapproachidentifying,
title={A joint optimization approach to identifying sparse dynamics using least squares kernel collocation},
author={Alexander W. Hsu and Ike Griss Salas and Jacob M. Stevens-Haas and J. Nathan Kutz and Aleksandr Aravkin and Bamdad Hosseini},
year={2025},
eprint={2511.18555},
archivePrefix={arXiv},
primaryClass={stat.ME},
url={https://arxiv.org/abs/2511.18555},
}MIT