NeurDE (pronounced nɜrd) is a teaching library with Jupyter notebooks for understanding some of the neural differential equations literature, namely
- Neural Ordinary Differential Equations (2018)
- Infinitely Deep Bayesian Neural Networks with SDEs (2021)
For the engineering bits, the codebase is as cogent and comprehensive as possible, without any cryptic optimization tricks, to better facilitate learning. As for the theoretical parts, the Jupyter notebooks contain a full dissection of neural ordinary differential equations (ODEs) and stochastic differential equations (SDEs).