报告题目:Towards Sustainable Distributed Machine Learning
报告人:徐平 助理教授 University of Texas Rio Grande Valley
报告时间:2024年7月15日(周一),下午4:30—6:30
报告地点:兴庆校区数学楼2-1会议室
报告摘要:Distributed machine learning has emerged as a trend to leverage abundant data and computational resources from geographically dispersed devices. By enabling local training on distributed devices and sharing only model parameters, this method ensures that private raw data remain on the device, thereby preserving privacy. However, distributed learning is still grappling significant challenges that hinder its practical application, including communication bottlenecks, privacy concerns, malicious attacks, and fairness issues. Addressing these challenges comprehensively is crucial for developing a sustainable machine learning system.
In this talk, I will highlight the motivation, challenges, and tasks involved in developing a sustainable machine learning framework over large-scale distributed networks. Our goal is to avoid raw data transmission over networks, improve communication efficiency, and enhance the trustworthiness of the developed machine learning framework. By focusing on these aspects, we aim to create a robust and efficient distributed learning system. The impact of the underlying distributed learning scheme on system performance, in terms of convergence behavior and generalization capability, will be thoroughly analyzed and demonstrated through both analytical and experimental results. This analysis will provide insights into optimizing distributed learning and making it more resilient to various issues, thereby contributing to its wider adoption and practical utility.
![](/__local/1/BA/71/2960F007C98A224C97361B32706_815B3839_10B64.png)
个人简介:Dr. Ping Xu is currently an Assistant Professor at the Electrical and Computer Engineering department of University of Texas Rio Grande Valley. Previously, she was a postdoctoral researcher with the Department of Electrical and Computer Engineering, George Mason Uni-versity, Fairfax, Virginia. Her research interests span the areas of machine learning and optimization, signal processing, dynamical systems, and cooperative control. She received the award of the Rising Star in EECS in 2022, the Outstanding Academic Achievement Award at GMU in 2022, and the IEEE Signal Processing Society Professional Development Grant in 2021.