In MultKAN forward():

1- postacts = postacts_numerical + postacts_symbolic

2- output_range = torch.std(postacts, dim=0) # for visualization, include the contribution from both spline + symbolic

3- self.edge_actscale.append(output_range)

In MultKAN reg():

elif reg_metric == 'edge_forward_spline_u':

`acts_scale = self.edge_actscale`

so "'edge_forward_spline_u`: the “norm” of edge, unnormalized (output std), only consider the spline (ignoring symbolic)" is wrong?

to correct in forward():
self.edge_actscale.append(output_range_spline)