Research Highlights

Jiayu's research is anchored in the vision of Unified Knowledge Integration - an innovative approach that harnesses AI to unify insights from large-scale, noisy, multimodal, and heterogeneous datasets. By seamlessly integrating domain knowledge with novel machine learning and generative AI techniques, we aim to transform various real-world domains, enhance decision making for data science through establishing the closed-loop flow of informatics among key components of human, data, and analytics. Please visit ILLIDAN Lab for projects, publication and research activities.

Machine Learning and AI+X Foundation: Knowledge Transfer and Integration

Jiayu has been working in the area of transfer learning which integrates domain knowledge from multiple sources to improve the generalization performance of machine learning models. His research focuses on multi-task learning, i.e., performing inductive knowledge transfer between multiple related learning tasks and improving the generalization performance simultaneously for all different tasks. Jiayu is the author of multi-task learning open-source software MALSAR.

Machine Learning and AI+X Foundation: Fusion of Information from Multiple Sources

In health informatics, complex underlying factors often contribute to events of interest, such as mortality, disease progression, and readmission. These events can be correlated to various variables from different sources. Jiayu's lab has been developing information fusion approaches that integrate multiple data sources/views as well as domain knowledge and information diffusion algorithms that transfer predictive knowledge between related analytics tasks.

Machine Learning and AI+X Foundation: Consolidating Data from Different Locations

A main obstacle to data analytics in the field of health informatics is the scattered nature of medical data, which is often dispersed across various locations. Distributed medical data presents critical challenges. Jiayu's lab has developed a suite of federated learning algorithms that provide privacy-preserving and fair predictive modeling from multiple collaborating parties with heterogeneous data distributions.

AI+Health: Early Diagnosis of Dementia

Jiayu has developed advanced disease progression models for Alzheimer’s, integrating data from electronic medical records (EHR) and brain imaging. These models effectively tackle challenges such as data sparsity, inconsistency, and high-dimensional features by combining multiple imaging modalities and leveraging insights from other neurodegenerative diseases. Recently, Jiayu's team explore how to leverage large-language models (LLMs) to improve early diagnosis of Alzheimer’s. Check out the preprint on exciting and surprising results based on a study using EHR from 2.5 million patients.

AI+Health: Language Markers and Chatbots for Dementia

One major research direction of Jiayu is to create digital biomarkers from conversations and daily behavior, offering affordable and accessible tools for early dementia screening. His team has also developed a chatbot to quickly gather language markers, significantly enhancing early-stage dementia prediction. Recently, his team is developing a chatbot that provides therapeutic treatment using Large-Language Models.

AI+Health: Drug Discovery

Jiayu's lab has been developing deep-learning models that can predict new therapeutic drug candidates by utilizing existing drug profiles from related disease models and identifying new drug targets. Jiayu's lab developed data-driven molecular fingerprint that leverages domain knowledge and enables cost-effective drug discovery through the virtual screening of chemical compounds, and a framework to search new molecules with desired properties. His lab is working on multi-modal foundation models that enables collaborative drug discovery with human experts.

Teaching

Jiayu is developing new curriculums at both undergraduate and graduate levels, that incorporate the state-of-the-art machine learning research into classroom.

• MSU CSE 480 [Spring 2022] [ Syllabus | Piazza | D2L ]
• MSU CSE 847 Machine Learning [Spring 2022] [ Syllabus | Piazza | D2L ] [Prev. Spring 2021 - 2016]
• MSU CSE 491/404 Introduction to Machine Learning [Prev. Fall 2020 - 2016]

Selected Publications

For recent preprints please check out the publication list of ILLIDAN Lab. For the full publication list see Jiayu's Google Scholar.

• Ranking Policy Gradient.
  Kaixiang Lin and Jiayu Zhou.
  ICLR 2020. Preprint 2019.
• Retaining Privileged Information for Multi-Task Learning.
  Fengyi Tang, Cao Xiao, Fei Wang, Jiayu Zhou, and Li-Wei Lehman.
  KDD 2019.
• Efficient Large-Scale Fleet Management via Multi-Agent Deep Reinforcement Learning.
  Kaixiang Lin, Renyu Zhao, Zhe Xu and Jiayu Zhou.
  KDD 2018. [Paper]
• Identify Susceptible Locations in Medical Records via Adversarial Attacks on Deep Predictive Models.
  Mengying Sun, Fengyi Tang, Jinfeng Yi, Fei Wang and Jiayu Zhou.
  KDD 2018. [Paper]
• Subspace Network: Deep Multi-Task Censored Regression for Modeling Neurodegenerative Diseases.
  Mengying Sun, Inci M. Baytas, Liang Zhan, Zhangyang Wang and Jiayu Zhou.
  KDD 2018. [Paper]
• Enhancing Predictive Modeling of Nested Spatial Data through Group-Level Feature Disaggregation.
  Boyang Liu, Pang-Ning Tan, and Jiayu Zhou.
  KDD 2018. [Paper]
• Multi-Modality Disease Modeling via Collective Deep Matrix Factorization.
  Qi Wang, Mengying Sun, Liang Zhan, Paul Thompson, Shuiwang Ji and Jiayu Zhou.
  KDD 2017 [Paper]
• Privacy-Preserving Distributed Multi-Task Learning with Asynchronous Updates.
  Liyang Xie, Inci Baytas, Kaixiang Lin and Jiayu Zhou.
  KDD 2017 [Paper]
• Patient Subtyping via Time-Aware LSTM Networks.
  Inci Baytas, Cao Xiao, Xi Zhang, Fei Wang, Anil Jain and Jiayu Zhou.
  KDD 2017 [Paper]
• Multi-Task Feature Interaction Learning.
  Kaixiang Lin, Jianpeng Xu, Inci M. Baytas, Shuiwang Ji and Jiayu Zhou.
  KDD 2016 [Paper]
• A Safe Screening Rule for Sparse Logistic Regression.
  Jie Wang, Jiayu Zhou, Jun Liu, Peter Wonka, Jieping Ye.
  NIPS 2014 [Paper]
• From Micro to Macro: Data Driven Phenotyping by Densification of Longitudinal Electronic Medical Records.
  Jiayu Zhou, Fei Wang, Jianying Hu, Jieping Ye.
  KDD 2014 [Paper] [Code].
• FeaFiner: Biomarker Identification from Medical Data through Feature Generalization and Selection.
  Jiayu Zhou, Zhaosong Lu, Jimeng Sun, Lei Yuan, Fei Wang, Jieping Ye.
  KDD 2013 [Paper] [Supplemental]
• Modeling Disease Progression via Fused Sparse Group Lasso.
  Jiayu Zhou, Jun Liu, Vaibhav A. Narayan, and Jieping Ye.
  KDD 2012 [Paper] [Code] Best Video Award [Info]
• Clustered Multi-Task Learning via Alternating Structure Optimization.
  Jiayu Zhou, Jianhui Chen and Jieping Ye.
  NIPS 2011 [Paper] [Code]
• A Multi-Task Learning Formulation for Predicting Disease Progression.
  Jiayu Zhou, Lei Yuan, Jun Liu and Jieping Ye.
  KDD 2011 [Paper] [Code]

Service

Jiayu serves as an Associate Editor for ACM Transactions on Computing for Healthcare, Neurocomputing. He and Journal of Alzheimer's Disease, and a Guest Editor for EURASIP Journal on Bioinformatics and Systems Biology and EURASIP Journal on Advances in Signal Processing. Jiayu is a dedicated peer reviewer for leading journals, including the Journal of Machine Learning Research (JMLR), IEEE Transactions on Knowledge and Data Engineering (TKDE), IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), and IEEE Transactions on Neural Network Learning Systems (TNNLS), among others. His contributions extend to serving on the Technical Program Committee (TPC) and organizing committees for numerous top-tier conferences such as KDD, ICML, NIPS, ICLR, IJCAI, AAAI, and more, where he has held key leadership roles, including SPC, Program Vice-Chair, and Workshop Chair.

Some words keep me moving forward

A job well done is its own reward. You take pride in the things you do, not for others to see, not for the respect, or glory, or any other rewards it might bring. You take pride in what you do, because you're doing your best. If you believe in something, you stick with it. When things get difficult, you try harder.