南京大学计算机科学与技术系
软件新技术与产业化协同创新中心
摘 要：
Recently,
the security in distributed machine-learning systems has drawn wide attention
in the community. Especially, an increasing amount of work has been conducted
on the problem of Byzantine attacks/failures, which assume the worst cases for
the distributed systems.
In
this talk, Xie is
going to introduce the recent progress in Byzantine-tolerant distributed
machine learning. He will briefly survey the 2 categories: robust statistics
and score-based approaches, and focus on his recent publications in this field.
报告人简介:
Cong Xie is a
Ph.D. candidate in University of Illinois Urbana Champaign, who expects to
graduate in 2021, advised by Prof. Indranil
Gupta and Prof. Oluwasanmi Koyejo. His
research focuses on protecting distributed learning algorithms from malicious
attacks, and designing communication-efficient distributed optimization
algorithms. He has presented his research results at leading conferences such
as ICML, NeurIPS,
UAI, ECML, etc. Xie was selected as a J.P. Morgan
Fellow (2020) for his work in secure distributed machine learning
时间：10月30日 10:00-11:30
腾讯会议平台ID: 266 171 407
|
