My name is Zhong Li, and I am a postdoc in Computer Science at Leiden University, affiliated with the EDA Lab. Before this, I did my PhD with Dr. Matthijs van Leeuwen (daily supervisor, promotor) and Prof. Dr. Thomas Bäck (promotor) at Leiden University. During my PhD, I worked as a visiting PhD researcher in the DAML group at Technical University of Munich in 2024, supervised by Prof. Dr. Stephan Günnemann.

Research Interest

I am broadly interested in the field of Data Mining and Machine Learning, especially trustworthy AI.

For my postdoc, I focus on trustworthy generative models: “when the generative models are able to provide insights into the underlying mechanisms of the data they generate, we should also provide such insights on the models themselves.” I am working on some topics like:

• Understanding of Unsupervised Learning: Why Self-Supervised Learning works (or not)? How to improve it?
• Diffusion Models: How to better extend diffusion models for tabular and/or discrete data? and What are they actually learning?

During my PhD, I work on trustworthy anomaly detection, with a focus on but not limited to complex data such as event sequence and graph-structured data for smart manufacturing. More concretely, I aim to improve the following aspects of anomaly detection methods:

• Accuracy: make anomaly detection models more accurate, especially with high-dimensional, unstructured, multimodal data;
• Explainability: make anomaly detection models more understandable for humans;
• Generalizability: make anomaly detection models work reliably under changing conditions such as data distribution shifts, uncharted hyperarameter configurations, and the presence of adversarial perturbations.

Recent News

• 2024.11: I will give a poster demonstration of our paper Explainable Graph Neural Networks Under Fire at the the Amsterdam meetup for the global LoG Conference.
• 2024.09: 🔥🔥 Our paper titled Cross-domain Graph Level Anomaly Detection is accepted by TKDE for publication!
• 2024.01: 🔥🔥 Our paper titled Graph Neural Networks based Log Anomaly Detection and Explanation is accepted by ICSE’24 poster track!
• 2023.07: 🔥🔥 Our paper titled A survey on explainable anomaly detection is accepted by TKDD for publication!
• 2023.07: 🔥🔥 Our paper titled Explainable Contextual Anomaly Detection using Quantile Regression Forests is accepted by DMKD for publication!
• 2022.12: 🔥🔥 Our paper titled Feature Selection for Fault Detection and Prediction based on Event Log Analysis is accepted by SIGKDD Explorations for publication!
• 2021.10: 🔥🔥 Our paper titled Choosing and changing the analysis scale in non-inferiority trials with a binary outcome is accepted by Clinical Trials for publication!
• [Position-Cancelled]: I am looking for a Master student to do an internship&master thesis, which will be in collaboration with Canon. The thesis proposal is “Knowledge-based & Data-driven Inkjet Jet Failure Detection and Classification”. (October 2023)
• [Position-Filled]: I am looking for a Master student to do an internship&master thesis, which will be in collaboration with VDL & ASML. The thesis proposal is “Predictive Maintenance of Lithography System Using Log Anomaly Detection”. (June 2023)

Professional Services

• Invited Reviewer for the following journals:
  • IEEE Transactions on Knowledge and Data Engineering (TKDE)
  • Data Mining and Knowledge Discovery (DMKD or DAMI)
  • International Journal of Computer Vision (IJCV)
  • IEEE Transactions on Emerging Topics in Computational Intelligence (TETCI)
  • IEEE Internet of Things Journal (IoT)
  • IEEE Intelligent Systems
  • AI Communications
• Programme Committee (PC) member for ECMLPKDD, and Reviewer for KDD, ICLR.

Contact

Office BE 2.23, Gorlaeus Building, Einsteinweg 55, 2333 CC Leiden, The Netherlands

Email: z.li(at)liacs.leidenuniv.nl