Adaptive Laplacian eigenfunctions as bases for regression analysis
Lei Ding, Xiaole Bai · Proceedings - International Conference on Pattern Recognition/Proceedings/International Conference on Pattern Recognition · 2008
Regression or least squares fitting is an important problem in statistics, data mining and many other applications. In recent years, basis functions derived from the underlying geometry of data, primarily Laplacian eigenfunctions, have attracted much interest. In this paper, we present a new framework based on adaptive Laplacian eigenfunctions and show the benefit of using a time-varying basis in regression analysis.