Wavelet neural networks generalization improvement

Similar to neural networks, the generalization improvement of wavelet neural networks is also an important issue since a given network may have good approximation accuracy, but could not perform well on unseen data. Generally, to improve generalization different techniques could be used including regularization. In this paper, two newly regularization techniques, applied to radial wavelet neural networks, are investigated. In the first technique, the additional term of the cost function is represented in terms of a Hilbert square norm of the functional representing the network structure. In the second technique however, the network adjusted parameters decay approach is used. Applied to wavelet neural networks, this type of approach includes all the adjusted parameters and not only the weights as with neural networks.