Methods for Structured Pruning of Neural Networks Norm of filter weights(L1, L2, .... Ln) Taylor approximation - Activation and gradient of activation wrt cost function Taylor approximation - Weight and gradient of weight wrt cost function