一、時間

2023年12月21日（周四）14：00

二、地點

學術交流中心201會議室

三、主要内容

報 告 主 題： Jointly Designing Communication Protocols and Model Update Compression in  Federated Learning

主    講    人 ：Dr Yipeng Zhou

主講人介紹：Dr Yipeng Zhou is a senior lecturer with the School of Computing, Faculty of Science and Engineering, Macquarie University. Before joining Macquarie University, he was a research fellow with the University of South Australia, and a lecturer with Shenzhen University, respectively. He got his Ph.D. and M.Phil degrees from The Chinese University of Hong Kong, and B.S. degree from University of Science and Technology of China, respectively. He received 2023 Macquarie University Vice-Chancellor's Research Excellence Highly Commended Award and 2023 IEEE Open Journal of the Communications Society Best Editor Award. He was the recipient of 2018 Australia Research Council Discover Early Career Researcher Award (DECRA). His research interests lie in federated learning, data privacy-preservation, networking, etc. He has published 110+ papers in top venues, including IEEE INFOCOM, 

背景介紹：Federated learning empowers geographically decentralized clients to collaboratively train a  machine learning model through exchanging model updates with a central server via Internet communications. However, massively transmitting model updates between the server and a large number of decentralized clients via Internet can be extremely resource consuming, making the training of advanced models in federated learning impracticable. In this talk, we will introduce our contributions to the design of model compression algorithms with the objective to minimize communication cost in federated learning without compromising model utility. More specifically, we jointly consider the application layer Internet protocol design and compression algorithm design so as to maximize model utility subjecting to limited communication resource. The experimental results conducted on real datasets demonstrate that our algorithms can considerably diminish communication cost in federated learning, especially when training high-dimension advanced models.