Importance Estimation

Filter pruning is an efficient way to structurally remove the redundant parameters in convolutional neural network, where at the same time reduces the computation, memory storage and transfer cost. Recent state-of-the-art methods globally estimate …