Boosting Kernel Density Estimation: Theory and Application

 

This talk will draw together two topics: Boosting - a method of classification first proposed within Machine Learning - and Kernel Density Estimation, which has also been used in discrimination problems. By applying boosting to kernel methods we show that boosting is a method of bias reduction, and enjoys similar properties to other such methods. Two kernel boosting algorithms are introduced - for density estimation, and for classification, and these are tested on simulated and real data with encouraging results.