AI-enhanced numerical solvers

Updated! TODO!

By Zhongzhan Huang, Mingfu Liang, Shanshan Zhong and Liang Lin.

The official implementation of the technical report paper "AttNS: Attention-Inspired Numerical Solving For Limited Data Scenarios" [paper]

🥳🥳AttNS has been accepted by ICML'24!

---

By Zhongzhan Huang, Senwei Liang, Hong Zhang, Haizhao Yang and Liang Lin.

The official implementation of our paper "On Fast Simulation of Dynamical System with Neural Vector Enhanced Numerical Solver" [paper]

🥳🥳NeurVec has been accepted by Scientific Reports!

Introduction

NeurVec is an open-source and data-driven corrector, which can break through the speed-accuracy trade-off of the large-scale simulations for dynamical systems. NeurVec can be easily plugged into the existing numerical solver, e.g. Euler methond, Runge–Kutta method, etc.

Requirement

• Python 3.7
• PyTorch 1.11

Dataset and pre-train model

• Dateset [Google] [Baidu] (verification code: g1pd)
• Pre-train model [Google] [Baidu] (verification code: sixu)

Citation

If you find this paper helps in your research, please kindly cite

@article{huang2023fast,
  title={On fast simulation of dynamical system with neural vector enhanced numerical solver},
  author={Huang, Zhongzhan and Liang, Senwei and Zhang, Hong and Yang, Haizhao and Lin, Liang},
  journal={Scientific Reports},
  volume={13},
  number={1},
  pages={15254},
  year={2023},
  publisher={Nature Publishing Group UK London}
}

Acknowledgement

We would like to thank zhengdao chen for his pytorch version of SRNN and Travis for his/her tutorial.