Clustering Based Gaussian Process Regression

Gaussian Process (GP) has become a common Bayesian inference framework and has been applied in many tasks, for example, data mining and non-linear transformation, in recent years. After introducing the model of GP regression and its training process, this thesis points out the inefficiency of the training process of GP, and proposes the modification to the original GP to speed up the training process by the clustering data set. Extensive experiments were conducted including model selection experiments and comparison experiments. The results show that the proposed algorithms are about 10 times faster than the original GP, with comparable precision