3 ms·

> It has been training that approximator. Also it is representation of approximator itself and data compression. For cubic splines to approximate some NN you l

> It has been training that approximator. Also it is representation of approximator itself and data compression. For cubic splines to approximate some NN you likely would need enormous number of intervals covering input space.

VHRanger 6y ago
Also, cubic splines don't extrapolate like the other mentioned models