An Exchanging-based Refinement to Sparse Gaussian Process Regression

We propose a backward deletion procedure to Sparse Gaussian Process Regression (SGPR) model, which can be used to refine a number of se- quential forward selection algorithms addressed recently. Some