Entropy Induced Pruning Framework for Convolutional Neural Networks
This addresses a time-consuming bottleneck in model compression for image classification tasks, though it is incremental as it builds on existing structured pruning methods.
The paper tackles the problem of structured pruning for convolutional neural networks being sensitive to poor training by proposing a metric called Average Filter Information Entropy (AFIE) to measure filter importance, achieving stable results even with models trained for only one epoch.
Structured pruning techniques have achieved great compression performance on convolutional neural networks for image classification task. However, the majority of existing methods are weight-oriented, and their pruning results may be unsatisfactory when the original model is trained poorly. That is, a fully-trained model is required to provide useful weight information. This may be time-consuming, and the pruning results are sensitive to the updating process of model parameters. In this paper, we propose a metric named Average Filter Information Entropy (AFIE) to measure the importance of each filter. It is calculated by three major steps, i.e., low-rank decomposition of the "input-output" matrix of each convolutional layer, normalization of the obtained eigenvalues, and calculation of filter importance based on information entropy. By leveraging the proposed AFIE, the proposed framework is able to yield a stable importance evaluation of each filter no matter whether the original model is trained fully. We implement our AFIE based on AlexNet, VGG-16, and ResNet-50, and test them on MNIST, CIFAR-10, and ImageNet, respectively. The experimental results are encouraging. We surprisingly observe that for our methods, even when the original model is only trained with one epoch, the importance evaluation of each filter keeps identical to the results when the model is fully-trained. This indicates that the proposed pruning strategy can perform effectively at the beginning stage of the training process for the original model.