Thesis Title: Automatic learning of wavelet networks: application to image classification

Abstract:

There is currently an exponential growth in research related to the training field of neural and wavelets networks. Despite the various training algorithms which had been suggested concerning progress achieved in the field, a radical solution has not been yet achieved. In this thesis we have contributed our findings to the theoretical study of wavelet networks. We proposed a new architecture of Hybrid beta wavelets networks ,as well as new algorithms of training of these networks. We then showed the capacities of the suggested hybrid beta wavelets networks(PBWN-OLS and FBWN) regarding a vast field of applications; that of image classification. This new approach of image classification proved to be precise on a broad range of problems.

Key-words: Beta hybrid wavelets network, learning algorithm, images classification