ELLIS Distinguished Lecture: Yizhou Yu

On December 4, 2026, Yizhou Yu (University of Hong Kong) will give a lecture: Progress in Neural Architectures for Computer Vision. Hosted by Guoying Zhao (University of Oulu and ELLIS Institute Finland).

Date and time 

December 4, 2026, 11:00-12:00 EET

Location

TU1 Saab Auditorium, Aalto University (Maarintie 8, Espoo)
Zoom link

Abstract

In the past decade, we have witnessed remarkable leaps in visual learning, from neural architectures and representation learning to foundation models. In this talk, we introduce a series of deep neural architectures for vision tasks. We first propose a pure ConvNet architecture that incorporates top-down attention, an important mechanism in the human vision system. Next, we present SegMAN, a linear-time semantic segmentation network supporting omni-scale context modeling with state space models and local attention. Then, we introduce nnFormer, a U-Net style 3D transformer for volumetric medical image segmentation. Finally, we present an analysis of artifacts in vision transformers, and discuss a solution named LaST-ViT that selectively integrates patch features into image representation.

Bio

Yizhou Yu is a chair professor in School of Computing and Data Science, and an associate director of the Institute of Data Science at the University of Hong Kong. He was also a faculty member at University of Illinois, Urbana-Champaign for more than ten years. He received his PhD in computer science from UC Berkeley. He has served on the editorial board of TPAMI, TIP and TVCG, and the program committee of many leading conferences. He is an ACM Fellow and IEEE Fellow. His current research interests include computer vision, machine learning, foundation models, and AI for medicine.