ThiNet: A Filter Level Pruning Method for Deep Neural Network Compression. We propose an efficient and unified framework, namely ThiNet, to simultaneously accelerate and compress CNN models in both training and inference stages. We focus on the filter level pruning, i.e., the whole filter would be discarded if it is less important. Our method does not change the original network structure, thus it can be perfectly supported by any off-the-shelf deep learning libraries. We formally establish filter pruning as an optimization problem, and reveal that we need to prune filters based on statistics information computed from its next layer, not the current layer, which differentiates ThiNet from existing methods. Experimental results demonstrate the effectiveness of this strategy, which has advanced the state-of-the-art. We also show the performance of ThiNet on ILSVRC-12 benchmark. ThiNet achieves 3.31× FLOPs reduction and 16.63× compression on VGG-16, with only 0.52% top-5 accuracy drop. Similar experiments with ResNet-50 reveal that even for a compact network, ThiNet can also reduce more than half of the parameters and FLOPs, at the cost of roughly 1% top-5 accuracy drop. Moreover, the original VGG-16 model can be further pruned into a very small model with only 5.05MB model size, preserving AlexNet level accuracy but showing much stronger generalization ability.
Keywords for this software
References in zbMATH (referenced in 5 articles )
Showing results 1 to 5 of 5.
- Ding, Xiaofeng; Yang, Hongfei; Chan, Raymond H.; Hu, Hui; Peng, Yaxin; Zeng, Tieyong: A new initialization method for neural networks with weight sharing (2021)
- He, Juncai; Jia, Xiaodong; Xu, Jinchao; Zhang, Lian; Zhao, Liang: Make (\ell_1) regularization effective in training sparse CNN (2020)
- Jang, Jinhyeok; Cho, Hyunjoong; Kim, Jaehong; Lee, Jaeyeon; Yang, Seungjoon: Deep neural networks with a set of node-wise varying activation functions (2020)
- Zheng, Qinghe; Tian, Xinyu; Yang, Mingqiang; Wu, Yulin; Su, Huake: PAC-Bayesian framework based drop-path method for 2D discriminative convolutional network pruning (2020)
- Zhang, Yunxiang; Blusseau, Samy; Velasco-Forero, Santiago; Bloch, Isabelle; Angulo, Jesús: Max-plus operators applied to filter selection and model pruning in neural networks (2019)