Open jbrea opened 6 years ago
Hmm, in theory, the NES solvers we do include should have the same performance. In practice, our NES implementations are less tested and my gut feeling is they are not yet "optimal". But please try the DX-NES solver and see if it can handle your problems.
Maybe @alyst has used the dxnes more and can share his experiences. Me, I tend to more often use the de methods since they tend to scale better to large dimensions but YMMV.
Are there any plans to include the popular cma-es solver? I am particularly interested in the option with noisy objectives...