Leveraging Locality and Robustness to Achieve Massively Scalable Gaussian Process Regression

Neural Information Processing Systems 

This has motivated extensive work on computationally efficient approximations.