W375 Westgate Building
1:30pm
In this talk, we present an importance-driven cross-layer design for semantic communication and networking. Building on statistical decision theory, we introduce an information-theoretic measure of per-sample data importance that quantifies the task-specific value of each data sample. Specifically, the importance, or value, of each data sample is characterized by the reduction in task-specific loss achieved by observing that sample. Using this metric, we formulate a cross-layer optimization problem that jointly optimizes (i) physical-layer joint source-channel coding (JSCC) and (ii) MAC-layer resource allocation, with the objective of maximizing semantic spectrum efficiency, defined as the data value delivered per unit bandwidth per unit time. At the physical layer, we develop the Meta-Learning Variational Information Bottleneck (Meta-VIB), a semantic communication transceiver that uses a compact neural network model with only 4.16 million parameters to generalize across varying signal-to-noise ratio (SNR), codelength, and Age of Information (AoI) values without requiring online retraining. At the MAC layer, we formulate channel allocation as a Multi-Action Restless Multi-Armed Bandit (MA-RMAB) problem and develop the Q-Maximization algorithm for dynamic channel resource allocation among sensors. Experimental results on a real-world pedestrian safety dataset collected at Toomer’s Corner in Auburn, Alabama, demonstrate that the proposed cross-layer design achieves substantial gains in semantic spectrum efficiency over baseline methods. We will also introduce SafeStep (https://safestep.eng.auburn.edu), a live, browser-based interactive platform for demonstrating semantic communication in pedestrian safety monitoring.
Additional Information:
BIOGRAPHY
Yin Sun is the Bryghte D. and Patricia M. Godbold Endowed Associate Professor in the Department of Electrical and Computer Engineering at Auburn University, Alabama. He received his B.Eng. and Ph.D. degrees in Electronic Engineering from Tsinghua University in 2006 and 2011, respectively. From 2011 to 2017, he was a Postdoctoral Scholar and Research Associate at The Ohio State University. He joined Auburn University as an Assistant Professor in 2017 and was promoted to Associate Professor in 2023. His research interests include Age and Semantics of Information, wireless networks, AI for 6G wireless systems, agriculture, education, and robotics. Dr. Sun has served on the editorial boards of the IEEE/ACM Transactions on Networking, IEEE Transactions on Information Theory, IEEE Transactions on Network Science and Engineering, IEEE Transactions on Green Communications and Networking, and the Journal of Communications and Networks. He has also served on the organizing committees of numerous international conferences, including as Technical Program Committee Chair for ACM MobiHoc 2025 and General Chair for IEEE/IFIP WiOpt 2026. He founded the Age and Semantics of Information (ASoI) Workshop in 2018 and the Modeling and Optimization in Semantic Communications (MOSC) Workshop in 2023. His publications have received multiple recognitions, including the Best Student Paper Award at IEEE/IFIP WiOpt 2013, the Best Paper Award at IEEE/IFIP WiOpt 2019, runner-up for the Best Paper Award at ACM MobiHoc 2020, the Best Paper Award from the Journal of Communications and Networks in 2021, the IEEE Communications Society William R. Bennett Prize in 2025, and the IEEE INFOCOM 2026 Test-of-Time Paper Award. He received the Auburn Author Award in 2020 and the National Science Foundation (NSF) CAREER Award in 2023.
Details...
