A Study on Gaussian Process Based Sequential Regression Models

Abstract

The c-regression models is a regression-based clustering method that requires a pre-defined number of clusters. The sequential regression models (SRM) is a method that extracts clusters sequentially without setting the number of clusters. Gaussian process based c-regression models(GPCRM) is a nonlinear and highly expressive c-regression models. In this study, we propose a Gaussian process based sequential regression models. The proposed method presents nonlinear regression models and automatically estimates the number of clusters. We evaluate the performance of the proposed method on two datasets with a nonlinear cluster structure. Experimental results show that the proposed method obtains a lower ARI than GPCRM and SRM.