Some Issues in Constructing Physics-Informed Neural Networks Based on KANs

[hzy24]

1. Nodes cannot directly fit multiplication and division operations; nodes can only use addition and subtraction.
2. When using the model, it is not possible to restrict the activation functions used for fitting. For example, I would like to add lib=["desired function to fit"] in the model configuration like model = KAN(width=[1, 2, 1], grid=5, k=3, device=device, symbolic_enabled=False).

[AlessandroFlati]

Please give a look to #123