キーワード: 正則化、一般回帰、不連続
We extend a regularization theory which is based on general regression in order to apply it to multi-valued regularization. This regularization theory do not require any smoothness measures which play an important role in standard regularization theory. We also propose a new discontinuous function reconstruction algorithm from discrete data without conventional line processes. This algorithm is very simple and leads deterministically to a unique solution. We show it's usefulness and robustness against noise by several computer simulations.
Keywords: regularization, general regression, discontinuity