A Novel Hybrid Computational Intelligence Approach for Solving General Constrained Global Optimization Problems

Abstract

Listen

Translate article

This study presents a novel hybrid computational intelligence method, which combines an improved quantumbehaved particle swarm optimization (IQPSO) method and a parallel multi-layer perceptron neural network (PMLPNN), namely IQPSO-PMLPNN algorithm, for solving general constrained global optimization (CGO) problems. The proposed IQPSO-PMLPNN algorithm has three features. The first, each decision variable and each penalty parameter for a constraint of a CGO problem are independently related to an individual network topology of a MLPNN. The second, each constraint has its optimal value of penalty parameter by using an evolutionary process. The third, a normalized and a renormalized operators are used to increase the capabilities of exploration and exploitation for decision variables and penalty parameters with their own search space. The proposed method can obtain a global optimal solution for a test general CGO problem by using the IQPSO approach to optimize the weights of network topology which consisting of decision variables and penalty parameters for a PMLPNN. Moreover, the performance of the proposed approach was compared with those of published algorithms, which individual genetic algorithms, artificial immune algorithms and hybrid computational intelligence (CI) methods. The numerical results indicate that the IQPSO-PMLPNN algorithm can find a global optimal solution for each test CGO problem, and that the numerical results obtained using the proposed algorithm are identical to those obtained using the hybrid CI approaches. The proposed approach can thus be considered as an efficient global optimization algorithm.