Nonlinear function estimation with locality adaptive feature weighting and its application to system identification

Just-In-Time(JIT) modeling is a predicting method which predicts the nonlinear func- tion from the locally linearized functions. What is important is that JIT modeling is under the strong influence of the metric which can he adjusted by feature wezghting. In this paper, we prp pose a new feature weighting method named locality adaptive feature weighting and demonstrate its effect on improving prediction accuracy.