Flexible and efficient Gaussian process models

I will briefly describe our work on the sparse pseudo-input Gaussian process (SPGP), where we refine the sparse approximation by selecting `pseudo-inputs' using gradient methods. I will then describe