{"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/accelerating-deep-unsupervised-domain","title":"Accelerating Deep Unsupervised Domain Adaptation with Transfer Channel Pruning","arxiv_id":"1904.02654","date":"2019-03-25","proceeding":null,"authors":["Chaohui Yu","Jindong Wang","Yiqiang Chen","Zijing Wu"],"abstract":"Deep unsupervised domain adaptation (UDA) has recently received increasing\nattention from researchers. However, existing methods are computationally\nintensive due to the computation cost of Convolutional Neural Networks (CNN)\nadopted by most work. To date, there is no effective network compression method\nfor accelerating these models. In this paper, we propose a unified Transfer\nChannel Pruning (TCP) approach for accelerating UDA models. TCP is capable of\ncompressing the deep UDA model by pruning less important channels while\nsimultaneously learning transferable features by reducing the cross-domain\ndistribution divergence. Therefore, it reduces the impact of negative transfer\nand maintains competitive performance on the target task. To the best of our\nknowledge, TCP is the first approach that aims at accelerating deep UDA models.\nTCP is validated on two benchmark datasets-Office-31 and ImageCLEF-DA with two\ncommon backbone networks-VGG16 and ResNet50. Experimental results demonstrate\nthat TCP achieves comparable or better classification accuracy than other\ncomparison methods while significantly reducing the computational cost. To be\nmore specific, in VGG16, we get even higher accuracy after pruning 26% floating\npoint operations (FLOPs); in ResNet50, we also get higher accuracy on half of\nthe tasks after pruning 12% FLOPs. We hope that TCP will open a new door for\nfuture research on accelerating transfer learning models.","url_abs":"http://arxiv.org/abs/1904.02654v1","url_pdf":"http://arxiv.org/pdf/1904.02654v1.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":"accelerating-deep-unsupervised-domain","repo_url":"https://github.com/jindongwang/transferlearning","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"domain-adaptation","task_name":"Domain Adaptation"},{"task_slug":"transfer-learning","task_name":"Transfer Learning"},{"task_slug":"unsupervised-domain-adaptation","task_name":"Unsupervised Domain Adaptation"}],"methods":[{"method_slug":"pruning","method_name":"Pruning"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}