Advancing Ubiquitous Intelligence Through Collaborative Learning and Intelligent Data Trading (video)

In the evolving landscape of data trading, the need for advanced methods to handle dynamic and diverse datasets has become increasingly crucial. This presentation introduces a novel framework for optimizing portfolio allocations in data trading markets using Autonomous Economic Agents (AEAs). Leveraging federated learning (FL), AEAs collaboratively train models while maintaining data privacy and ownership. Our approach addresses the challenges of dynamic revenue returns by integrating Histogram of Oriented Gradients (HoG) with Discrete Wavelet Transformation (DWT), enhancing the representation of non-stationary revenue patterns. Additionally, we propose a frequency-domain transformation method to efficiently manage model parameter drifts, reducing communication overhead and improving training efficiency. This presentation outlines the proposed data trading market, the innovative techniques for handling dynamic data, and the benefits of our approach in terms of privacy, data quality, and scalability. Bio: Lei Zhao received the B.S. and M.A.Sc. degrees in computer science and technology from Xidian University, Xi'an, China, in 2015 and 2018, respectively, and earned his Ph.D. in Electrical and Computer Engineering from the University of Victoria in 2023. He is currently a Post-Doctoral Fellow and Sessional Lecturer in the E&CE Department at the University of Victoria. His research focuses on federated learning and optimization with applications in finance.