| Learning from Examples with Quadratic Mutual Information (2000) | |||||||||||||||||
Abstract | |||||||||||||||||
| This paper discusses a novel algorithm to train nonlinear mappers with information theoretic criteria (entropy or mutual information) directly from a training set. The method is based on a Parzen window estimator and uses Renyi's quadratic definition of entropy and a distance measure based on the Cauchy-Schwartz inequality. We apply the algorithm to the difficult problem of vehicle pose estimation in synthetic aperture radar (SAR) with very good results. 1.0 Introduction Information theory has served a crucial role in communication theory, but its application to pattern recognition and information processing has been sporadic. At the core lies the difficulty that pattern recognition is a discipline based on the learning by example metaphor, while information theory principles require an analytic form for the probability density function (pdf). One possibility is to postulate the form of the pdfs (normally Gaussian) and estimate from the data their parameters (mean and variance for Ga... | |||||||||||||||||
Details der Publikation | |||||||||||||||||
| |||||||||||||||||