Sun-Yuan Kung

Sun-Yuan Kung is a Professor of Electrical and Computer Engineering at Princeton University specializing in high-performance learning networks and explainable AI. His research develops beyond back-propagation paradigms to create optimal neural architectures through structural learning and discriminant information metrics.

His academic credentials include:

• Ph.D. in Electrical Engineering, Stanford University (1977)
• M.S. in Electrical Engineering, University of Rochester (1974)
• B.S. in Electrical Engineering, National Taiwan University (1971)

Professor Kung pioneered Explainable Neural Networks (XNN) incorporating Internal Neuron's Learnability (INL) using internal teacher labels and discriminant information (DI) for node/layer ranking. This enables deep compression while enhancing model robustness and adaptability for real-time applications in image processing, privacy protection, and security-sensitive environments. His work directly addresses DARPA's Explainable AI (XAI) initiatives through end-user-adaptive labeling mechanisms.

Recent publications (2023-2025) demonstrate sustained innovation in efficient deep learning architectures, with dominant themes in remote sensing applications (object detection, image captioning, change detection), quantum-inspired algorithms, and multimodal representation learning. Key trends include neural architecture search, model compression techniques, and reinforcement learning integration for specialized domains.

His distinguished honors include:

• Life Fellow of IEEE (2016)
• IEEE Third Millennium Medal (2000)
• IEEE Signal Processing Society Best Paper Award (1996)
• IEEE Signal Processing Society Technical Achievement Award (1992)
• IEEE Fellow (1988)

Professor Kung mentors graduate researchers including Katherine Shu-Min Li, Xiaxin Shen, and Zhuoqing Song, with funding supporting projects in deep compression, privacy-preserving ML, and XAI frameworks. His group develops foundational techniques for discriminant information-based network pruning and structural gradient methods.

He leads a research team advancing the XNN framework for DARPA's Explainable AI initiatives, focusing on real-time decision support through internal neuron explainability and channel exploration for model optimization.