- Re-visit PhyLSTM with code from paper authors [priority:: low] [completion:: 2025-08-01]
https://github.com/zhry10/PhyLSTM/blob/master/model/PhyLSTM3.py
The code from the original authors is available, and it turns out Francesco did not write his code from the author repository, but rather from his interpretation of the PhyLSTM paper. This could explain why the model does not converge on the training data, and diverges when L-BFGS is used at the end of training.