Structure optimization of wavelet neural network using rough set theory

This paper presents an approach to minimize the redundancy of structure existing in frame-based wavelet neural networks using the rough sets theory. The original structure of the wavelet network is obtained through a time-frequency analysis. Then the redundant nodes are eliminated in light of the dependency between the output of the network and nodes in the hidden layers to optimize the structure of the wavelet network. Simulation results show the proposed method is simple and effective.