George K. Thiruvathukal
Affiliation: Loyola University Chicago
Country: USA
Title: Reusing Deep Learning Models: Trust, Challenges, and Directions for Software Engineering 2.0 in an Era of Generative AI
Abstract
Deep neural networks (DNNs) achieve state-of-the-art performance in many areas, including computer vision, natural language processing, system configuration, and question-answering. However, DNNs are expensive to develop, both in intellectual effort (e.g., devising new architectures) and computational costs (e.g., training). Reusing DNNs offers a promising solution to amortize these costs, both within organizations and across the broader computing industry. As the paradigm shifts towards Software 2.0, however, understanding reuse patterns becomes critical. Trust in deep learning models is essential, particularly as they become integrated with traditional (non-ML) software. Software engineering must also evolve to account for the growing use of DNN components, including generated artifacts. In the era of generative AI, where models and prompts serve as valuable artifacts, preserving them is crucial for ensuring reproducibility. This invited talk describes the challenges in current DNN reuse, including both technical gaps and missing engineering practices. We summarize failures in reuse techniques across conceptual (e.g., reuse based on research), adaptive (e.g., building on existing implementations), and deployment (e.g., reuse on new devices) approaches, while outlining advances needed to improve each.
Bio
George K. Thiruvathukal is Full Professor of Computer Science and the Department Chairperson at Loyola University Chicago. He is also Visiting Computer Scientist at Argonne National Laboratory in the Leadership Computing Facility. He received a Ph.D. (1995) and MS (1990) in Computer Science from Illinois Institute of Technology and a BA (1988) in Computer Science and Physics with a Mathematics Minor from Lewis University in Romeoville, IL. As a Computer Science major at Lewis University, he received the department’s top graduating student award in the College of Arts and Sciences; as a Physics major at the same university, he was inducted into the Sigma Pi Sigma Physics National honor society. During the summer of his junior year, he worked with Philip J. Hatcher at the University of New Hampshire on a NSF-funded summer REU program focused on compiler construction for data-parallel languages on genearl-purpose MIMD architectures, where he really learned C and Unix systems programming and continues to teach on these topics to this day in courses such as COMP 141, 310-410, and 339-439 (among others). The resulting tools led to the development of a portable Data Parallel C dialect inspired by the C* Language used to program the Connection Machine. The ideas of Data Parallel C live on in Data Parallel C++, an Intel effort where Dr. Thiruvathukal is developing curricular modules to bring this promising new technology to new audiences.