Sparse Multi-output Gaussian Processes

In this work we propose a sparse approximation for the full covariance matrix involved in the multiple output convolution process. We exploit the fact that each of the outputs is conditional independe