tonisunset

Arxiv is not peer-reviewed and usually used for preprint versions before submitting a paper to a journal. Also it allows to make the information accessible for free to anybody.

This is interesting and differs from other ODE ML approaches in that it can deal with non-conservative forces, like friction. The design allows for a shallow network with only 3 free parameters and is therefore very cheap to train.

yes, try fitting a function to less than pi/4 period of data...