BYY Harmony Learning on Weibull Mixture with Automated Model Selection

Abstract

Bayesian Ying-Yang (BYY) harmony learning has provided a new learning mechanism to implement automated model selection on finite mixture during parameter learning with a set of sample data. In this paper, two kinds of BYY harmony learning algorithms, called the batch-way gradient learning algorithm and the simulated annealing learning algorithm, respectively, are proposed for the Weibull mixture modeling based on the maximization of the harmony function on the two different architectures of the BYY learning system related to Weibull mixture such that model selection can be made automatically during the parameter learning on Weibull mixture. The two proposed algorithms are both demonstrated well by the simulation experiments on some typical sample data sets with certain degree of overlap.