Pruning product unit neural networks | IEEE Conference Publication | IEEE Xplore