报告题目：Mathematical Foundation of Distributed Machine Learning

报告人：吴强教授，Middle Tennessee State University

报告时间：2023年12月11日（周一）9:30-10:30

报告地点：腾讯会议ID：437-564-841

报告摘要：Machine learning is a central component of data science and plays increasing roles in modern industries. The successful applications of machine learning algorithms in practice usually motivate theoretical studies on their computational and mathematical properties, which help researchers and practitioners to better understand the algorithms, identify appropriate application domains, and set up hyperparameters to achieve the best performance. This, however, is only one side of the story. On the other side, theoretical studies can also in turn motivate new algorithms by addressing the limitations of existing algorithms. This usually improves the performance in some specific scenarios or broadens the application domains of existing algorithms. In this talk, I will present the mathematical foundations of the divide and conquer approach for distributed machine learning. We proved the minimax optimality of several distributed kernel regression approaches. Based on these studies, we designed a bias correction strategy to improve the performance of distributed kernel regression and a recentering regularization approach to make distributed learning applicable in other machine learning tasks.

报告人简介：吴强教授于2005年于香港城市大学获得博士学位，2005-2008年在美国杜克大学进行博士后研究，之后先后在美国密歇根州立大学、英国利物浦大学、美国中田纳西州立大学工作。吴强教授的研究方向集中于统计学习理论及其应用。吴强教授已经发表论文80余篇，并获得包括美国自然科学基金、美国农业部、Simons基金会等多项资助。

邀请人：向道红