While the research on convolutional neural networks (CNNs) is progressing quickly, the real-world deployment of these models is often limited by computing resources and memory constraints. In this paper, we address this issue by proposing a novel filter pruning method to compress and accelerate CNNs. Our method reduces the redundancy in one convolutional layer by applying sparse subspace clustering to its output feature maps. In this way, most of the representative information in the network can be retained in each cluster. Therefore, our method provides an effective solution to filter pruning for which most existing methods directly remove filters based on simple heuristics. The proposed method is independent of the network structure, and thus it can be adopted by any off-the-shelf deep learning libraries. Evaluated on VGG-16 and ResNet-50 using ImageNet, our method outperforms existing techniques before fine-tuning, and achieves state-of-the-art results after fine-tuning.
展开▼
