Theme：Recent advance in time series sensor data analytics

Reporter：Prof. Xiaoli Li

Prof. Xiaoli Li, Director and Chief Scientist at the Institute for Infocomm Research (I²R), Agency for Science, Technology and Research (A*STAR), Singapore, Adjunct Professor at the School of Computer Science and Engineering, Nanyang Technological University (NTU), Singapore. He is an IEEE Fellow, Fellow of the Asia-Pacific Artificial Intelligence Association (AAIA), recognized as one of the World's Top 2% Scientistsin artificial intelligence by Stanford University, and a Highly Cited Researcher(cross-field) by Clarivate.

Time: 10-11am, May 16, 2025

Venue: Room 5008, Building 2, iHarbour

Abstract:

The widespread deployment of sensors across industries such as manufacturing, aerospace, and healthcare has generated vast amounts of time series data, creating an urgent demand for advanced AI-driven analytics. This talk presents recent breakthroughs in AI techniques designed to unlock the full potential of sensor-based time series data. We will highlight progress in three key areas: (1) self-supervised representation learning, which leverages contrastive learning to extract rich features from unlabeled time series data; (2) unsupervised domain adaptation for multivariate sensor streams, addressing local and global distribution shifts to ensure robust cross-domain performance; and (3) model compression and optimization for efficient edge deployment, enabling real-time analytics in resource-constrained environments. Finally, we will explore emerging efforts in developing foundation models for time series analytics, and how they can be effectively adapted to diverse downstream applications.