时       间：2024年5月30日 上午9:00 – 10:30

地       点：华中科技大学东五楼210学术报告厅

主       题：Rethinking Benchmarks for Machine Learning Systems

嘉       宾：何丙胜 新加坡国立大学  教授  计算机学院副院长

报告摘要：

The advent of new machine learning applications presents a unique opportunity to rethink and redesign our benchmarking approaches. This involves re-evaluating how we measure machine learning system performance and abstracting various design factors for these systems. In this presentation, we will discuss our efforts in developing benchmarks tailored to different machine learning applications. First, we have built a benchmark based on distinct design principles for the Non-IID data distributions in federated learning. This benchmark, NIID-Bench (https://github.com/Xtra-Computing/NIID-Bench), is specifically designed to assess model accuracy and other critical aspects of federated learning systems. Second, we have developed a benchmark focusing on real-time feature extractions for online applications, FEBench (https://github.com/decis-bench/febench). Our benchmark studies have yielded numerous intriguing results and insights, particularly in benchmarking machine learning systems. Finally, we will highlight some of the open challenges in creating benchmarks for future machine learning systems.

报告人简介:

Dr. Bingsheng He is currently a Professor and Vice-Dean (Research) at School of Computing, National University of Singapore. Before that, he was a faculty member in Nanyang Technological University, Singapore (2010-2016), and held a research position in the System Research group of Microsoft Research Asia (2008-2010), where his major research was building high performance cloud computing systems for Microsoft. He got the Bachelor degree in Shanghai Jiao Tong University (1999-2003), and the Ph.D. degree in Hong Kong University of Science & Technology (2003-2008). His current research interests include cloud computing, database systems and high-performance computing. He has been a winner for industry faculty awards from Microsoft/Google/NVIDIA/Xilinx/Alibaba. His work also won multiple recognitions as “Best papers” collection or awards in top forums such as SIGMOD 2008, VLDB 2013 (demo), IEEE/ACM ICCAD 2017, PACT 2018, IEEE TPDS 2019, FPGA 2021 and VLDB 2023 (industry). Since 2010, he has (co-)chaired a number of international conferences and workshops, including IEEE CloudCom 2014/2015, BigData Congress 2018, ICDCS 2020 and ICDE 2024. He is an ACM Distinguished member (class of 2020).

主       题：面向新兴计算架构的高效图数据处理

嘉       宾：孙世轩  上海交通大学 长聘教轨副教授

报告摘要：

作为有效建模和分析实体间关联关系的方式，图被广泛用于社交网络、在线支付、互联网等实际应用中。然而，图数据的海量性、稀疏性和异构性，以及图计算负载的多重动态性，为大规模图计算的性能和硬件资源的有效利用带来巨大挑战。为了应对上述挑战，我们着重研究面向新兴计算架构的图数据处理，基于图数据和计算负载特性，挖掘新兴计算架构的优势，提升系统的高效性。本次报告将介绍我们在基于Serverless架构和GPU加速的图数据处理方面的进展。

报告人简介:

孙世轩博士目前是上海交通大学计算机科学与工程系长聘教轨副教授。此前，于香港科技大学获得博士学位（2015-2020），并在新加坡国立大学从事博士后研究员工作（2020-2023）。他的主要研究方向是大数据系统和并行计算，目前专注于高性能图数据处理的研究；研究成果发表在SIGMOD、VLDB、ASPLOS、ICDE等顶级会议。他入选了国家级青年人才引进计划，上海市青年人才引进计划等项目。