Note that KANs use LBFGS, which is second-order optimization method. My experience with the use of second-order features suggests that simple gradient descent o

Note that KANs use LBFGS, which is second-order optimization method. My experience with the use of second-order features suggests that simple gradient descent often leads to divergence.