Convergence suppression and divergence facilitation: new approach to prune hidden layer and weights of feedforward neural networks | IEEE Conference Publication | IEEE Xplore