* 이번 학기는 온라인으로 열리게 되며 학부생, 대학원생, 교직원 누구나 참여 가능합니다. 

세미나 6 – 정보통신대학원 세미나 2021년 1학기

Program Repair meets Deep Learning

김동선 교수 (경북대학교)

초록

Fixing program bugs are often tedious and time-consuming. Many software projects are, however, relying on manual debugging. Automated program repair has been studied in several different directions to deal with the current practice. One of the most promising research lines is pattern-based program repair. Many studies have already leveraged open-source projects as a large number of debugging cases are available in those projects. Since manual mining is not scalable, it is necessary to adopt an automated approach. Recent advancements of deep representation learning provide opportunities to automate pattern mining for program repair. This talk presents the results of two studies using deep repair pattern mining. The first study explores a space of fix patterns for static analysis violations; the study proposes an approach to leveraging deep neural networks to identify common fix patterns that resolve alarms found by FindBugs. The second study presents how to use deep feature representation techniques for debugging method names. In the study, we investigate a large number of pairs of method names and their implementations to identify inconsistent pairs. Based on the results, we propose a novel technique to identify and refactor inconsistent method names by using deep neural networks.

Bio: Dongsun Kim is an assistant professor at Kyungpook National University. He has received his Ph.D. degree in Computer Science and Engineering from Sogang University, Korea. His career includes several academic and industrial experiences. He was a postdoctoral fellow at the Hong Kong University of Science and Technology from September 2010 to June 2013. He joined the University of Luxembourg as a research associate in November 2013 and continued his position until November 2018. From April 2019, he worked as a senior software test engineer position at Furiosa.ai, a fabless startup company manufacturing neural processing units.

He has published several research papers and participated in several research projects relevant to AI-based software engineering. In particular, he has pioneered a new line of research on pattern-based program repair. His recent achievements have focused on automated fix pattern mining, deep code representation for mining fix patterns, program repair driven by bug reports, fault localization impact on program repair, and specific topics for program repair; for these topics, he leveraged several AI-based techniques such as autoencoders, CNNs, etc. He has led a relevant research project, “Automated Program Repair using Fix Patterns Learned from Human-written Patches” (499,000EUR), funded by Luxembourg National Research Fund (FNR) as sole-PI. He also participated in another relevant project as co-PI, “Automated Fixing of Program Vulnerabilities in the Android Ecosystem” (127,000EUR), funded by the University of Luxembourg.

[대학원]세미나(5월 26일, 김동선 교수) – Program Repair meets Deep Learning
Post Views: 54