Experiments on the CEC 2015 expensive optimization testbed

We experiment various simple classical algorithms on the expensive optimization testbed from Cec2015. CMA performs best, in particular its DCMA flavor using quasi-random numbers. Nelder-Mead also performs well. Portfolios performed well for the given budget (500 evaluations in dimension 10 and 1500 evaluations in dimension 30), but not the half budget, which is also taken into account in the competition, hence we did not include them in the final version.