Accelerating the training of feedforward neural networks using generalized Hebbian rules for initializing the internal representations | IEEE Journals & Magazine | IEEE Xplore