In this paper, we propose a novel progressive parameter pruning method for Convolutional Neural Network acceleration, named Structured Probabilistic Pruning (SPP), which effectively prunes weights of convolutional layers in a probabilistic manner. Unlike existing deterministic pruning approaches, where unimportant weights are permanently eliminated, SPP introduces a pruning probability for each weight, and pruning is guided by sampling from the pruning probabilities. A mechanism is designed to increase and decrease pruning probabilities based on importance criteria in the training process. Experiments show that, with 4x speedup, SPP can accelerate AlexNet with only 0.3% loss of top-5 accuracy and VGG-16 with 0.8% loss of top-5 accuracy in ImageNet classification. Moreover, SPP can be directly applied to accelerate multi-branch CNN networks, such as ResNet, without specific adaptations. Our 2x speedup ResNet-50 only suffers 0.8% loss of top-5 accuracy on ImageNet. We further show the effectiveness of SPP on transfer learning tasks.
@article{arxiv.1709.06994,
title = {Structured Probabilistic Pruning for Convolutional Neural Network Acceleration},
author = {Huan Wang and Qiming Zhang and Yuehai Wang and Haoji Hu},
journal= {arXiv preprint arXiv:1709.06994},
year = {2018}
}
Comments
CNN model acceleration, 13 pages, 6 figures, accepted by Proceedings of the British Machine Vision Conference (BMVC), 2018 oral