Jialei Wang

Right now the optimization routines that maximize ExpectedImprovement are all designed for continuous space unfortunately. A simple fix would be to add optimization methods that work for discrete space, for...

If you only search along one of the dimensions while the other is fixed, you should consider using 1-dimension GP instead of 2-d, because the dimension which is fixed is...

Interesting, you are right. In this case you should use 2-d GP for a better modeling. The current implementation of MOE assumes the search dimension is the same as the...