Simple, basic implementation of neural ODE in autograd/numpy. For learning only
This implements the ideas in the paper "Neural Ordinary Differential Equations" in as simple a form as possible, using only autograd. It is not efficient. It is not useful for any practical purpose. Use torchdiffeq for any real use.
[1] Ricky T. Q. Chen, Yulia Rubanova, Jesse Bettencourt, David Duvenaud. "Neural Ordinary Differential Equations." Advances in Neural Processing Information Systems. 2018. [arxiv]
The implementation is based on the write up of Per Vognsen in terms of the costate vector, which is a clear exposition of how the adjoint method works.