An Analysis of Bayesian Bagging Prediction for Improving Generalization Performance
Shuichi Kurogi, Kenta Harashima, Nozomi Shibata · International Conference on Intelligent Information Processing · 2010
This paper presents an analysis of a method to improve the generalization performance of bagging predictions by means of Bayesian approach. We show a formalization of Bayesian prediction using bagging machines for regression problems, and present a method to reduce the generalization loss. We examine and analyze to show the usefulness of the present method via numerical experiments using bagging CAN2 as a bagging machine, where the CAN2 is a neural net for learning efficient piecewise linear approximation of nonlinear functions.