Abstract:

Interpreting and debugging machine learning models is necessary to ensure the robustness of the machine learning models. Explaining mispredictions can help significantly in doing so. While recent works on misprediction explanation have proven promising in generating interpretable explanations for mispredictions, the state-of-the-art techniques “blindly” deduce misprediction explanation rules from all data features, which may not be scalable depending on the number of features. I will present our research on alleviating this problem, which proposes an efficient misprediction explanation technique leveraging prior knowledge about data. I will also present our work reducing model’s bias by focusing on subsets of data during training that are more prone to being mispredicted.

Bio:

 

Iftekhar Ahmed received his B.Sc. degree in Computer Science and Engineering from Shahjalal University of Science and Technology, Bangladesh. He received Ph.D. degree in Computer Science from Oregon State University. He is an Assistant Professor in the School of Information and Computer Sciences, University of California, Irvine, where he leads the Software Engineering & Testing Using Artificial Intelligence for Reliable Software (STAIRS) group. He is also a faculty member of the Institute for Software Research, UCI. His research interests include the field of software engineering, with a focus on combining testing, static analysis, socio-technical factors analysis, and machine learning approaches to help improve software quality under realworld conditions. His Ph.D. work on improving the effectiveness of mutation analysis for largescale real-world software systems has helped to identify several bugs in the Linux kernel. The improvements resulting from his work have been incorporated into the Linux distributions, with more than 2 Billion instances running worldwide from mobile phones to data centers. His recent work on test smells detection is implemented as a plugin for PyCharm, a popular IDE for Python developed by JetBrains with more than 10 million developers using it worldwide. He is a member of the ACM, the ACM SIGSOFT, and the IEEE.

 

Abstract:
Can a Large Language Model generate appropriate commit messages based solely on code diffs? This talk describes a study my students and I did on this topic, highlighting the potential, and some limitations, of LLMs for this particular software engineering task.

Bio:
Cristina (Crista) Lopes is a Professor in the School of Computer Sciences at University of California, Irvine, with research interests in Programming Languages, Software Engineering, and Distributed Virtual Environments. She is an IEEE Fellow, an ACM Distinguished Scientist, a twice-elected member of the SIGPLAN Executive Committee, and founding Editor in Chief of The Art, Science, and Engineering of Programming. She is the recipient of the 2016 Pizzigati Prize for Software in the Public Interest for her work in the OpenSimulator virtual world platform.

Abstract:
Open-source software has become increasingly popular in industrial products with the advent of open-source software. Due to several open-source software alternatives for a particular purpose, industry practitioners must decide among multiple candidates for adoption. One of the primary sources of information about open-source software is its documentation. Despite this, there are no studies regarding the importance of documentation in OSS adoption decisions. Therefore, our first step is to investigate the role of this software artifact in the industry's adoption decision-making process. After learning how the industry uses documentation when adopting, we will use natural language processing (NLP) techniques to extract the industry's preferred information from the documentation.

Bio:
Mohammad Moshirpour is the director of the Software Engineering Practice and Education (SEPE) Research Group. He is an associate professor of teaching at the Department of Informatics, Donald Bren School of Information & Computer Sciences at the University of California, Irvine. His research interests include software design and development, software requirements engineering, machine learning, and software engineering training and education. He is the recipient of several awards for his work in developing tools and methodologies for software engineering practices and education including the 2021 D2L Innovation Award.

Abstract:
Compositional AI systems, which combine multiple artificial intelligence components together with other application components to solve a larger problem, have no known pattern of development and are often approached in a bespoke and ad hoc style. This makes development slower and harder to reuse for future applications. To support the full rapid development cycle of compositional AI applications, we have developed a novel framework called (Bee)* (written as a regular expression and pronounced as "beestar"). We illustrate how (Bee)* supports building integrated, scalable, and interactive compositional AI applications with a simplified developer experience.

Bio:
Sergio Gago-Masague is an Assistant Professor of Teaching in Computer Science at UC Irvine, specializing in cross-disciplinary research involving software engineering, IoT and AI. His research focuses on designing and prototyping assistive and educational technologies in various fields, including health, energy efficiency, and pedagogy. His work has been recognized with support from prestigious funding sources, both private and public, including NIH and NSF. Sergio's teaching portfolio includes a range of courses such as data structures, the internet of things, capstone projects, and next generation search systems. Additionally, he serves as the Director of the Engaging Technology and Application Design (ETAD) Lab, the Center for Data-Driven Drugs Research and Policy (C3DRP) and the Capstone Project Program in Computer Science, where he leads the development, implementation, and assessment of AI-based learning strategies in project-based settings.