Modeling the Heterogeneity of Heterogeneity: 

Algorithms, Theories and Applications

Summary

Multiple types of heterogeneity naturally co-exist in a variety of high-impact real applications, such as manufacturing processes, abnormal user detection, etc. State-of-the-art data mining techniques are rich in handling a single type of heterogeneity, which are designed for a specific type of heterogeneity, and often inadequate in addressing the co-existence of multiple types of heterogeneity, referred to as Heterogeneity of Heterogeneity (HoH). To be specific, the following two questions largely remain unanswered: (Q1) how to jointly model multiple types of heterogeneity; (Q2) how to theoretically understand the model generalization performance in the context of HoH? The overall goal of this project is to model and understand the Heterogeneity of Heterogeneity in data mining, by (1) creating a suite of effective and efficient algorithms for modeling HoH, (2) characterizing the model generalization performance in the context of HoH , and (3) evaluate the proposed techniques on various real applications, such as abnormal user detection and quality control in manufacturing processes.

Publications

Tutorials

1. Jun Wu, Jingrui He. Trustworthy Transfer Learning: Transferability and Trustworthiness, KDD 2023. link

2. Jun Wu, Wenxuan Bao, Jingrui He. Learning from Non-IID Data: Centralized vs. Federated Learning, IJCAI 2023. link

3. Dongqi Fu, Zhe Xu, Hanghang Tong, Jingrui He. Natural and Artificial Dynamics in GNNs, WSDM 2023. link

4. Lecheng Zheng, Jingrui He. Contrastive Learning: A Heterogeneous Perspective, SDM 2023. link

5. Jun Wu, Wenxuan Bao, Jingrui He. Learning from Non-IID Data: Centralized vs. Federated Learning. IEEE BigData 2022. link

6. Dawei Zhou, Jingrui He. Gold Panning from the Mess: Rare Category Exploration, Exposition, Representation, and Interpretation. KDD 2019. link

7. Yao Zhou, Fenglong Ma, Jing Gao, Jingrui He. Optimizing the Wisdom of the Crowd: Inference, Learning, and Teaching. KDD 2019. link