The Matern 5/2 kernel was proposed in [1] as a good choice of GP kernel due to several advantageous properties. I propose to implement a GPMatern52EiTuner within btb that uses the Matern 5/2 kernel. (i.e. Matern(nu=5/2))
Upon completion, can delete the add-gpmatern52ei branch which contains an implementation on btb v0.2
[1]: Jasper Snoek, Hugo Larochelle, and Ryan P Adams. 2012. Practical bayesian optimization of machine learning algorithms. In NIPS. 2951– 2959.