| Summary: | The optimization problems and algorithms are the basics subfield in artificial intelligence, which is booming in the almost any industrial field. However, the computational cost is always the issue which hinders its applicability. This paper proposes a novel hybrid optimization algorithm for solving expensive optimizing problems, which is based on particle swarm optimization (PSO) combined with Gaussian process (GP). In this algorithm, the GP is used as an inexpensive fitness function surrogate and a powerful tool to predict the global optimum solution for accelerating the local search of PSO. In order to improve the predictive capacity of GP, the training datasets are dynamically updated through sorting and replacing the worst fitness function solution with the better solution during the iterative process. A numerical study is carried out using twelve different benchmark functions with 10, 20 and 30 dimensions, respectively. Regarding solving of the ill-conditioned computationally expensive optimization problems, results show that the proposed algorithm is much more efficient and suitable than the standard PSO alone.
|
|---|