Dr. Sun Yuan Kung

Biography

Sun-Yuan Kung focuses on developing high-performing learning networks and deep learning processors that have virtually replaced traditional signal processors for speech and image processing applications. His research emphasizes back-propagation as the de facto training paradigm for deep learning models, where useful parameter learning plays a crucial role in finding the optimal network structure. He has contributed to structural learning and the importance of node and layer ranking, which greatly enhance the process of network pruning, referred to as deep compression. Kung’s work includes developing joint parameter and structural gradient-type methods that gradually optimize network structures through a comprehensive solution space. Furthermore, he has introduced the Explainable Neural Network (XNN), incorporating internal teacher labels and internal optimization metrics, such as discriminant information, to facilitate effective internal learning paradigms. This research supports DARPA’s Explainable AI initiatives and addresses the challenges of real-time decision-making in active learning environments. His areas of expertise include Biological and Biomedical Computing, Networking, and Data and Information Science.

Research Interests