This paper presents a novel algorithm named HPSODE for constrained optimization problems. The proposed algorithm integrates particle swarm optimization (PSO) with differential evolution (DE) on the basis of an optimal information sharing mechanism firstly, which avoids premature convergence defects of the single algorithm. Then under the guidance of the feasibility rules, the algorithm quickly finds better feasible solution. Finally, HPSODE is tested on two engineering design problems. Comparisons show that HPSODE has higher computational precision, better robustness and is more effective for solving constrained optimization problem.
In this paper, a novel evolutionary functional network (EFN) is introduced. Firstly the generalized basis function is introduced, and then genetic programming is improved by changing the objects and structure of encoding. Sequences of generalized basis functions acts as individuals, general tree structure is used to encode them. Least square method (LSM) is used to design fitness function and by a number of evolutions, the optimum approximated model is achieved. The algorithm is used to compute the numerical integrals of all kinds of functions. Finally, results of 5 experiments show that this algorithm is effectively feasible and more accurate.