A Unifying View of Sparse Approximate Gaussian Process Regression

We provide a new unifying view, including all existing proper probabilistic sparse approximations for Gaussian process regression. Our approach relies on expressing the effective prior which the me...