joint work with Yichi Shen
We consider the optimization of multimodal, computationally expensive noisy functions f(x) with the assistance of a Radial Basis Function (RBF) surrogate approximation of f(x). This RBF surrogate is then used to help guide the optimization search to reduce the number of evaluations required. Because the function is noisy, it is necessary to resample some points to estimate the mean variance of f(x) at each location. The results compare favorably to Bayesian Optimization methods for noisy functions and can be more flexible in the choice of objective function.
joint work with Youhei Akimoto
We introduce an acceleration for covariance matrix adaptation evolution strategies (CMA-ES) by means of adaptive diagonal decoding. This diagonal acceleration endows the default CMA-ES with the advantages of separable CMA-ES without inheriting its drawbacks. Diagonal decoding can learn a rescaling of the problem in the coordinates within linear number of function evaluations. It improves the performance, and even the scaling, of CMA-ES on classes of non-separable test functions that reflect, arguably, a landscape feature commonly observed in practice.
This talk will discuss several improvements to the open-source library RBFOpt for surrogate model based optimization. RBFOpt is used in commercial applications and practical needs drove its development. We will discuss efficient model selection, local search, parallel optimization and - time permitting - multi objective optimization. Part of this talk is devoted to engineering issues, but we will try to be precise regarding the mathematics for efficient model selection and local search. This is based on results scattered in various recent papers; some details are currently unpublished.