Research on Knowledge-Guided Continual Learning

Associate Professor Xialei Liu
Nankai University, China

Continual learning is one of the key skills for the next generation of artificial intelligence systems, aiming to enable systems to continuously acquire new knowledge in dynamically changing environments while avoiding catastrophic forgetting of old knowledge, simulating human learning. This report will introduce, from the perspective of knowledge guidance, how to utilize model knowledge, attribute knowledge, textual knowledge, and external data knowledge to resist forgetting while enhancing knowledge transferability. A series of continual learning methods can be applied to various computer vision tasks, including image classification, image generation, image segmentation, and object detection. Finally, a summary and outlook on the progress in the field of continual learning will be provided.

Biosketch

Xialei Liu is currently an associate professor at the School of Computer Science, Nankai University, specializing in open-environment visual continual learning. He holds a Ph.D. from the Autonomous University of Barcelona, Spain, and worked as a postdoctoral researcher at the University of Edinburgh, UK. He engaged in machine learning and computer vision research in open environments, including continual learning, unsupervised learning, and few-shot learning. He has published over 30 academic papers, with over 3000 citations on Google Scholar, including top international journals and conferences such as TPAMI, NeurIPS, CVPR, and ICCV. One paper was selected as a CVPR 2022 Best Paper Finalist. He serves as a committee member of VALSE 2022-2024, and organized the CVPR 2023 Continual Learning Workshop.