中国计算机学会理论计算机科学,中国计算机学会

Yeh-Ching Chung香港中文大学深圳分校

Bio：Yeh-Ching Chung received a B.S. degree in Computer Science from Chung Yuan Christian University in 1983, and the M.S. and Ph.D. degrees in Computer and Information Science from Syracuse University in 1988 and 1992, respectively.  From 1992 to 2002, he was with the Department of Information Engineering and Computer Science at Feng Chia University, where he was an associate professor in 1992 and a full professor in 1999.  From 1993 to 1997, he served as the director of Computer Network Division of Computer Center. From 1998 to 2001, he was the chairman of the department.  From 2002 to 2016, he was with the Department of Computer Science at National Tsing Hua University as a full professor.  From 2003 to 2012, he served as the deputy director of Library, where he established the first UHF RFID library system in Taiwan.  In 2007, he founded Taiwan Association of Grid Computing (TAGC) that was renamed to Taiwan Association of Cloud Computing (TACC) in 2010.  He was the direct general of TAGC/TACC from 2007 to 2011.  He has supervised two National Tsing Hua University teams to win the champion of Student Cluster Competition, sponsored by IEEE/ACM SC conference, in 2010 and 2011.  He has served as General Chairs, Program Chairs, Keynote Speakers, and Technical Committee Members of many international conferences.  He was an awardee of Thousand Talents Plan of China in 2015.  In 2016, he joined the Laboratory of Cloud Computing and Disaster Recovery Technology at Research Institute of Tsinghua University in Shenzhen as a deputy director.  He is now a professor in School of Science and Engineering of Chinese University of Hong Kong in Shenzhen.  His research interests include parallel and distributed processing, cloud computing, big data, and embedded systems.  He has published over 200 journal and conference papers and developed many systems in these areas.

讲座题目：Graph Computing Technique for AI and Big Data Applications

摘要：In this talk, we will show how to apply the graph computing technique to two applications, association rule mining and large-scale software system optimization.  In association rule mining, the apriori method is the most commonly used approach to find the maximum frequent k-itemset. When the value of k is large, the apriori method is time consuming and it may not be able to get the result sometimes.  We have proposed a hybrid method, ANG, by combining the graph computing and apriori methods for the maximum frequent k-itemset calculation.  In ANG, when the value of k is small, the apriori method is used to calculate the maximum frequent k-itemset.  When k is over a threshold, the graph computing method is used to calculate the maximum frequent k-itemset. The experimental results show that ANG outperforms the apriori method for all test cases.

For the large-scale software system optimization, the main goal is to build up a program execution behavior monitoring and analyzing system. The system is an HPC system that integrates Cloud Computing, Distributed Storage, In-Memory Computing, Graph Computing, Compiler, Profiling Tools, and Data Mining techniques. When a large-scale software system is running, with the profiling tools, the execution behaviors of the software system can be recorded in real-time. With the real-time execution behavior records, different big data analytical methods can be used to optimize different desired parameters statically or dynamically. We have applied this system to optimized an LLVM compiler. The experimental results show that the proposed system has 5-10% performance improvement for the LLVM compiler.