Learning How To Localise Faults Automatically
Shin You is an associate professor in School of Computing at KAIST, Republic of Korea. He received a PhD in Computer Science from King’s College London, United Kingdom, and previously has been a lecturer at University College London, UK. His research interests include Search Based Software Engineering, i.e. the application of metaheuristic optimisation to software engineering problems, as well as evolutionary computation and genetic programming. He is the program co-chair of IEEE International Conference on Software Testing, Verification & Validation (ICST) 2018, and the Silver Medal recipient of ACM SIGEVO Human Competitiveness Award (HUMIES) 2017.
Automatic Fault Localisation is a technique that aims to identify the location of faults based on observations of test execution. After over a decade of research based on human
designed techniques, we are now treating the problem as a learning problem: given many examples of faults, can a machine learn how to tell the location of a fault from test executions?
We briefly look at the history of the field, and go through the recent theoretical analysis that dictates that the widely studied Spectrum Based Fault Localisation technique cannot produce
a single method that works best for all possible faults. Finally, the talk will present the latest results from machine learning based approach towards automated debugging that aims to
overcome the limits in SBFL.