LSU ECE Seminar Announcement and Schedule On Thursday, 10 April 2025, 8:30 - 9:30 (Public Portion), in Room 3316E Patrick F. Taylor Hall: Towards Utility of Unlabeled Data and Training Efficiency of Federated Learning a dissertation to be defended by Xiaobing Chen Ph.D. Candidate LSU Division of Electrical & Computer Engineering Abstract--This research focuses on advancing machine learning in resource-constrained settings by improving self-supervised learning for Human Activity Recognition (HAR) and developing robust federated learning frameworks. It introduces Temporal Contrastive Learning in Human Activity Recognition (TCLHAR), a method leveraging adjacent time windows to reduce labeled data reliance and better model dynamic processes. Besides, the work addresses federated learning efficiency by optimizing client selection and training procedures for diverse data and latency conditions. It also proposes a Multi-Group Transmission (MGT) scheme using OFDMA to reduce stochastic variance and accelerate model convergence, and finally explores incentive mechanisms that enhance both server and client utility in large-scale federated learning deployments. Thursday, 10 April 2025, 8:30 - 9:30 (Public Portion) Room 3316E Patrick F. Taylor Hall -------------------------------------------------------------------------------- Schedule as of 24 Mar 2025. For updates visit https://www.lsu.edu/eng/ece/seminar No other future seminars currently scheduled. For the latest schedule, abstracts of upcoming talks, and PDF announcements suitable for posting visit https://www.lsu.edu/eng/ece/seminar