题目: Creating Robust and Trustworthy Deep Learning Systems: A Software Quality Assurance Perspective
报告人：Prof. Jianjun Zhao（日本九州大学）
地点：第周苑 C-320（原N3楼 320）
Deep learning (DL) systems have achieved great success in many application domains such as image processing, speech recognition, and autonomous vehicles. Deep neural networks (DNNs) are the key driving force behind its recent success, but still seem to be a magic black box lacking interpretability and understanding. This brings up many open safety and security issues with enormous and urgent demands on rigorous methodologies and engineering practice for quality enhancement. Traditional software represents its logic as control flows crafted by human knowledge, while DNN characterizes its behaviors by the weights of neuron edges and the nonlinear activation functions (determined by the training data). Therefore, detecting erroneous behaviors in a DNN is different from that of traditional software in nature, which necessitates effective analysis, testing and verification techniques for DL systems.
In this talk, I will introduce several novel testing and analysis techniques for ensuring the safety and security of DL systems from a software quality assurance perspective, and point out research opportunities on creating robust and trustworthy DL systems.
Jianjun Zhao is a professor at the Faculty of Information Science and Electrical Engineering, Kyushu University, Japan. He received his B.E. degree in Computer Science from Tsinghua University in 1987, and Ph.D. degree in Computer Science from Kyushu University in 1997. He had been a visiting scientist in MIT Laboratory for Computer Science from April 2002 to March 2003. His main research interest is software engineering and programming language, in particular robust deep learning systems, program analysis and verification, automatic programming, software testing, debugging, and programming environments. He has published more than 100 research papers in conferences such as ICSE, ESEC/FSE, ASE, PLDI, ECOOP and ISSTA. He obtained the ACM SIGSOFT Distinguished Paper Award of ASE 2018 and the Best Paper Candidate Award of SANER 2016.