Title :
An Approach of Structural Design for Multi-resolution Wavelet Network Based on TBIC Rule
Author :
Yang, Huizhong ; Zhong, Hao
Author_Institution :
Southern YangtzUniversity, Wuxi
Abstract :
Based on testing Bayesian information criterion (TBIC), an approach was presented to solve the indeterminacy of structure existing in multi-resolution wavelet network. A method considering both scale and precision of the networks was proposed, which can also result in an optimal network structure under the TBIC rule. Simulation results illustrate that the multi-resolution wavelet network has reasonable scale and good testing performance when the method was applied. Compared with Bayesian information criterion (BIC), the TBIC rule can reflect accurately the trend of precision along with the structure.
Keywords :
Bayes methods; neural nets; wavelet transforms; multiresolution wavelet network; testing Bayesian information criterion; Artificial neural networks; Bayesian methods; Convergence; Design engineering; Electronic mail; Function approximation; Multiresolution analysis; System testing; Systems engineering and theory; Wavelet analysis;
Conference_Titel :
Natural Computation, 2007. ICNC 2007. Third International Conference on
Conference_Location :
Haikou
Print_ISBN :
978-0-7695-2875-5
DOI :
10.1109/ICNC.2007.181