Universität Bielefeld - Technische Fakultät

Stefan Klanke

I am no longer at Bielefeld University. For a current list of publications and software tools, please visit my personal page.

Unsupervised Kernel Regression

Read our paper "Principal Surfaces from Unsupervised Kernel Regression" (Meinicke, Klanke, Memisevic and Ritter)
in the September 2005 issue of IEEE Transactions on Pattern Analysis and Machine Intelligence.
IEEE Digital Library or Draft version

Abstract

We propose a nonparametric approach to learning of principal surfaces based on an unsupervised formulation of the Nadaraya-Watson kernel regression estimator. As compared with previous approaches to principal curves and surfaces the new method offers several advantages: first of all it provides a practical solution to the model selection problem, because all parameters can be estimated by leave-one-out cross-validation without additional computational cost. In addition, our approach allows for a convenient incorporation of nonlinear spectral methods for parameter initialization, beyond classical initializations based on linear PCA. Furthermore, it shows a simple way how to fit principal surfaces in general feature spaces, beyond the usual data space setup. The experimental results illustrate these convenient features on simulated and real data.


An outdated web page about sound synthesis done with UKR:
http://www.dspirit.de/ukr
Web Administration, 2005-07-07