A new class of quasi-Newtonian methods for optimal learning in MLP-networks | IEEE Journals & Magazine | IEEE Xplore