Globally Soft Filter Pruning For Efficient Convolutional Neural Networks
This paper propose a cumulative saliency based Globally Soft Filter Pruning (GSFP) scheme to prune redundant filters of Convolutional Neural Networks (CNNs).Specifically, the GSFP adopts a robust pruning method, which measures the global redundancy of the filter in the whole model by using the soft pruning strategy. In addition, in the model recovery process after pruning, we use the cumulative saliency strategy to improve the accuracy of pruning. GSFP has two advantages over previous works:(1) More accurate pruning guidance. For a pre-trained CNN model, the saliency of the filter varies with different input data. Therefore, accumulating the saliency of the filter over the entire data set can provide more accurate guidance for pruning. On the other hand, pruning from a global perspective is more accurate than local pruning. (2) More robust pruning strategy. We propose a reasonable normalization formula to prevent certain layers of filters in the network from being completely clipped due to excessive pruning rate.
PDF Abstract
MNIST
