A new learning algorithm for fully complex-valued RBF networks

作者：Liu, Shulin; Huang, He

来源：31st Youth Academic Annual Conference of Chinese Association of Automation, YAC 2016, Wuhan, China, 2016-11-11 To 2016-11-13.

DOI：10.1109/YAC.2016.7804940

• Complex-valued radial basis function network
• Generalization capability
• Inter-class distance
• Intra class
• Levenberg-Marquardt algorithm
• LM algorithm
• Radial basis function neural networks
• UCI machine learning repository

It has been known that the centers initialization and parameters updating algorithm are two crucial factors in radial basis function neural network (RBFNN) training process. This paper focuses on the learning of complex-valued radial basis FCRBF) networks. A distance-based center initialization method, where both the interclass distance and intraclass difference are taken into consideration, is proposed to determine the centers. Then, the complex-valued Levenberg-Marquardt (LM) algorithm is employed to adjust all the parameters of the FCRBF network. The performance of the proposed algorithm is evaluated using some datasets from UCI machine learning repository. The experimental results show that better generalization capability is achieved by the well-trained FCRBF network.

• 出版日期2016