Jiayi Shen (沈佳怡)

I am a final year ELLIS PhD candidate at MultiX Lab, University of Amsterdam. I am grateful to be supervised by Prof. Dr. Marcel Worring and Dr. Nanne van Noord. I also received valuable guidance from Dr. Xiantong Zhen. Prior to that, I obtained my Master's and Bachelor's degrees from Beihang University.

Email  /  CV  /  Google Scholar  /  Twitter  /  Github

profile photo
News

[Sep. 2024] One paper was accepted by NeurIPS 2024!

[June 2024] I am happy to give a talk "Probabilistic Modeling for Knowledge Transfer" for Yingzhen's group at Imperial College London.

[Feb. 2024] One paper on adapting foundation models was accepted by CVPR 2024.

[Feb. 2024] One tutorial on "How to Improve Your Academic Reading Skills" in MultiX. Slides can be found HERE!

Research

I'm interested in Machine Learning, Bayesian Models, and Meta-Learning. Much of my research is to address challenging problems in multi-task learning.

GO4Align: Group Optimization for Multi-Task Alignment
Jiayi Shen, Cheems Wang, Zehao Xiao, Nanne Van Noord, Marcel Worring
NeurIPS, 2024
paper / code

This paper proposes GO4Align, a multi-task optimization approach that tackles task imbalance by explicitly aligning the optimization across tasks.

Any-Shift Prompting for Generalization over Distributions
Zehao Xiao, Jiayi Shen, Mohammad Mahdi Derakhshani, Shengcai Liao, Cees Snoek,
CVPR, 2024
paper / code (Coming soon)

We propose any-shift prompting, a general probabilistic inference framework that considers both the training and test task information for prompt generation, to improve the generalization on downstream tasks across various distribution shifts by utilizing the original generalization ability.

Episodic Multi-Task Learning with Heterogeneous Neural Processes
Jiayi Shen, Xiantong Zhen, Qi (Cheems) Wang, Marcel Worring
NeurIPS, 2023, Spotlight
paper / code

This paper focuses on the data-insufficiency problem in multi-task learning within an episodic training set-up.

SuperDisco: Super-Class Discovery Improves Visual Recognition for the Long-Tail
Yingjun Du, Jiayi Shen, Xiantong Zhen, Cees Snoek
CVPR, 2023
paper / code

We propose SuperDisco to discover super-class representations for long-tailed recognition using a graph model.

Association Graph Learning for Multi-Task Classification with Category Shifts
Jiayi Shen, Zehao Xiao, Xiantong Zhen, Cees Snoek, Marcel Worring
NeurIPS, 2022
paper / code

We propose a new MTL setting which suffers from category shifts from training to test data.

NFormer: Robust Person Re-identification with Neighbor Transformer
Haochen Wang, Jiayi Shen, Yongtuo Liu, Yan Gao (Xiaohongshu Inc.), Efstratios Gavves
CVPR, 2022
paper / code

We propose a Neighbor Transformer Network, or NFormer, which explicitly models interactions across all input images.

Variational Multi-Task Learning with Gumbel-Softmax Priors
Jiayi Shen, Xiantong Zhen, Marcel Worring, Ling Shao
NeurIPS, 2021
paper / code

We propose variational multi-task learning (VMTL), a general probabilistic inference framework for learning multiple related tasks.

A Bit More Bayesian: Domain-Invariant Learning with Uncertainty
Zehao Xiao, Jiayi Shen, Xiantong Zhen, Ling Shao, Cees Snoek
ICML, 2021
paper / code

To better explore domain invariant learning, we introduce weight uncertainty to the model by leveraging variational Bayesian inference.


This website is based on Jon's source code.