Adaptive knot Placement for Nonparametric Regression
Hossein L. Najafi, Vladimir Cherkassky · Neural Information Processing Systems · 1993
Performance of many nonparametric methods critically depends on the strategy for positioning knots along the regression surface. Constrained Topological Mapping algorithm is a novel method that achieves adaptive knot placement by using a neural network based on Kohonen's self-organizing maps. We present a modification to the original algorithm that provides knot placement according to the estimated second derivative of the regression surface