Combined Electromagnetism-Like Mechanism Optimization Algorithm and ROLS with D-Optimality Learning for RBF Networks
The paper proposed a new self-constructed radial basis function network designing method via a two-level learning hierarchy. Aiming at getting stronger generalization ability and robustness, an integrated algorithm which combines the regularized orthogonal least square with learning with D-optimality experimental design method was introduced at the lower level, while electromagnetism-like mechanism algorithm for global optimization was employed at the upper level to search the optimal combination of three important learning parameters, i.e., the radial basis function width, regularized parameter and D-optimality weight parameter. Through simulation results, the effectiveness of the proposed algorithm was verified.
Keywordsradial basis function network two-level learning hierarchy electromagnetism-like mechanism algorithm generalization ability
Unable to display preview. Download preview PDF.
- 3.Chen, S., Hong, X., Harris, C.J.: Probability Density Function Estimation Using Orthogonal Forward Regression. In: Proceedings of International Joint Conference on Neural Networks, Orlando, Florida (2007)Google Scholar
- 5.Zhao, W.B., Yang, L.Y., Wang, L.M.: The Hybrid Structure Optimization Algorithms of Radial Basis Probabilistic Neural Networks. Acta Simulata Systematica Sinica 16(10), 2175–2184 (2004)Google Scholar
- 11.Chen, J.F., Ren, Z.W.: A New Two-level Learning Design Approach for Radial Basis Function Neural Network. Computer Simulation 26(6), 151–155 (2009)Google Scholar