A Novel Evolutionary Algorithm for Efficient Minimization of Expensive Black-box Functions with Assisted-Modelling

This work presents a novel algorithm for efficient global minimization of expensive black-box functions. A dedicated evolutionary algorithm is used to handle expensive and discontinuous functions; the EA also utilizes information from local-searches to efficiently bias its domain exploration. To enhance efficiency, the algorithm incorporates a density cluster analysis algorithm and a trust-region derivative-free optimizer. The algorithm performs well both when benchmarked against other candidate algorithms over a wide range of test functions and in a challenging real-world optimization problem.