{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/discrimination-aware-channel-pruning-for-deep","title":"Discrimination-aware Channel Pruning for Deep Neural Networks","arxiv_id":"1810.11809","date":"2018-10-28","proceeding":"NeurIPS 2018 12","authors":["Zhuangwei Zhuang","Mingkui Tan","Bohan Zhuang","Jing Liu","Yong Guo","Qingyao Wu","Junzhou Huang","Jinhui Zhu"],"abstract":"Channel pruning is one of the predominant approaches for deep model\ncompression. Existing pruning methods either train from scratch with sparsity\nconstraints on channels, or minimize the reconstruction error between the\npre-trained feature maps and the compressed ones. Both strategies suffer from\nsome limitations: the former kind is computationally expensive and difficult to\nconverge, whilst the latter kind optimizes the reconstruction error but ignores\nthe discriminative power of channels. To overcome these drawbacks, we\ninvestigate a simple-yet-effective method, called discrimination-aware channel\npruning, to choose those channels that really contribute to discriminative\npower. To this end, we introduce additional losses into the network to increase\nthe discriminative power of intermediate layers and then select the most\ndiscriminative channels for each layer by considering the additional loss and\nthe reconstruction error. Last, we propose a greedy algorithm to conduct\nchannel selection and parameter optimization in an iterative way. Extensive\nexperiments demonstrate the effectiveness of our method. For example, on\nILSVRC-12, our pruned ResNet-50 with 30% reduction of channels even outperforms\nthe original model by 0.39% in top-1 accuracy.","url_abs":"http://arxiv.org/abs/1810.11809v3","url_pdf":"http://arxiv.org/pdf/1810.11809v3.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"discrimination-aware-channel-pruning-for-deep","repo_url":"https://github.com/SCUT-AILab/DCP","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"model-compression","task_name":"Model Compression"},{"task_slug":"channel-selection","task_name":"channel selection"}],"methods":[{"method_slug":"pruning","method_name":"Pruning"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1810.11809","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}