{"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/improving-device-edge-cooperative-inference","title":"Improving Device-Edge Cooperative Inference of Deep Learning via 2-Step Pruning","arxiv_id":"1903.03472","date":"2019-03-08","proceeding":null,"authors":["Wenqi Shi","Yunzhong Hou","Sheng Zhou","Zhisheng Niu","Yang Zhang","Lu Geng"],"abstract":"Deep neural networks (DNNs) are state-of-the-art solutions for many machine\nlearning applications, and have been widely used on mobile devices. Running\nDNNs on resource-constrained mobile devices often requires the help from edge\nservers via computation offloading. However, offloading through a\nbandwidth-limited wireless link is non-trivial due to the tight interplay\nbetween the computation resources on mobile devices and wireless resources.\nExisting studies have focused on cooperative inference where DNN models are\npartitioned at different neural network layers, and the two parts are executed\nat the mobile device and the edge server, respectively. Since the output data\nsize of a DNN layer can be larger than that of the raw data, offloading\nintermediate data between layers can suffer from high transmission latency\nunder limited wireless bandwidth. In this paper, we propose an efficient and\nflexible 2-step pruning framework for DNN partition between mobile devices and\nedge servers. In our framework, the DNN model only needs to be pruned once in\nthe training phase where unimportant convolutional filters are removed\niteratively. By limiting the pruning region, our framework can greatly reduce\neither the wireless transmission workload of the device or the total\ncomputation workload. A series of pruned models are generated in the training\nphase, from which the framework can automatically select to satisfy varying\nlatency and accuracy requirements. Furthermore, coding for the intermediate\ndata is added to provide extra transmission workload reduction. Our experiments\nshow that the proposed framework can achieve up to 25.6$\\times$ reduction on\ntransmission workload, 6.01$\\times$ acceleration on total computation and\n4.81$\\times$ reduction on end-to-end latency as compared to partitioning the\noriginal DNN model without pruning.","url_abs":"http://arxiv.org/abs/1903.03472v1","url_pdf":"http://arxiv.org/pdf/1903.03472v1.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":"improving-device-edge-cooperative-inference","repo_url":"https://github.com/hou-yz/pytorch-pruning-2step","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null}],"tasks":[],"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}