A Study on Gaussian Process based c-Regression Models

Abstract

Gaussian process regression is a method to estimate Gaussian distributions of unknown outputs. c-regression models is a method to perform clustering and regression simultaneously. We propose Gaussian process based c-regression models (GPCRM). We show that the proposed method is theoretically equivalent to kernel c-regression models (KCRM). We compared GPCRM to KCRM using four artificial datasets and evaluated the cluster partition. The experiments also showed that the proposed method yields cluster partitions equivalent to KCRM. We showed that GPCRM is faster than KCRM.