A novel online adaptive mixed learning algorithm of the radial basis function neural network

作者：Zhai Huawei^*; Cui Licheng; Zhang Weishi

来源：Journal of Computational Information Systems, 2014, 10(4): 1405-1412.

DOI：10.12733/jcis9209

• Generalization capability
• Hidden nodes
• Online optimization algorithms
• Overlap factor
• Principle component analysis
• Radial basis function neural networks
• RBF Neural Network
• Sequential learning algorithm

Based on the importance of hidden nodes, this paper presents a novel online adaptive mixed learning algorithm of a RBF neural network, which combines the offline initializing method and online optimization algorithm. The algorithm uses the principle component analysis to initialize the network structure, so as to avoid wasting much time at the beginning of sequential learning algorithm, and it introduces the overlap factor, by using adding, deleting and updating hidden nodes, strictly controls the hidden node numbers and makes its structure optimum. After test comparing with other algorithms, it is superior in real-time, reliability and generalization capability.

• 出版日期2014