An Unsupervised Gaussian Mixture Classification Mechanism based on Statistical Learning Analysis

This paper presents a scheme for unsupervised classification with Gaussian mixture models by means of statistical learning analysis. A Bayesian Ying-Yang harmony learning system acts as a statistical tool for the particular derivation and development of automatic joint parameter learning and model selection. The proposed classification mechanism roughly decides on the number of real classes, by earning activation for the winners and assigning penalty for the rivals, so that the most competitive center wins for possible prediction and the extra ones are driven far away when starting the algorithm from a too large number of classes without any prior knowledge. Simulation experiments prove the feasibility of the approach and show good performance for unsupervised classification and natural estimation on the number of classes.