Introduction of LDSEE

LDSEE focus on expensive optimization, in which the objective is to find the global minimum of a given function within a very limited number of function evaluations. It draws much attention in recent years. Solving the expensive optimization problem by constructing a metamodel (or a series of metamodels, see Fig 1) has shown to be effective in practice. By metamodel we mean the approximate model of the original objective function and its constraints. 

Fig 1. A demo of metamodel-assisted optimization process

The present expensive optimization algorithms focus their attention on model approximation techniques, and use global optimization (GO) as black-box solver. So it is difficult for them to keep a good balance between model approximation and global search due to their two-part property. To overcome this difficulty, we try to unpack an efficient evolutionary algorithm, low dimensional simplex evolution (LDSE), and repack it with model approximation techniques in this paper. The proposed algorithm borrows ideas from tabu search and simulated annealing. It is inherently parallel and self contained. This renders it very easy to use. Numerical results show that our proposed algorithm is a competitive alternative for expensive optimization problems.

Fig 2. Idea of LDSEE

References

[1] C.T. Luo, S.-L Zhang, C. Wang, Z.L. Jiang, A metamodel-assisted evolutionary algorithm for expensive optimization, Journal of Computational and Applied Mathematics, 236 (2011) 759–764 
(DOI: 10.1016/j.cam.2011.05.047)