Open Akaqox opened 3 weeks ago
I understand now it means regularization and we can choose between
reg_metric : the regularization metric 'edge_forward_spline_n', 'edge_forward_spline_u', 'edge_forward_sum', 'edge_backward', 'node_backward' lamb_l1 : float l1 penalty strength lamb_entropy : float entropy penalty strength lamb_coef : float coefficient penalty strength lamb_coefdiff : float coefficient smoothness strength I hope that when you are free, you can fix the tutorial.
On Physics_2B_conservation_law_2D.ipynb tutorial notebook on the section:
reg_loss = model.reg(lamb_l1=1., entropy_offset=1e-4, lamb_coef=1.)
it says MultKAN.reg() got an unexpected keyword argument 'entropy_offset' and if I am remove this statement and retry as
it says MultKAN.reg() missing 3 required positional arguments: 'reg_metric', 'lamb_entropy', and 'lamb_coefdiff'
I dont know how to handle these arguments and what is the best values for this tutorial problem.