[VU ML Seminar] [VUML Seminar] Next meeting

[Yang, Z. (Zhao)]

Dear all,

Please see the information of our next VUML seminar (https://sites.google.com/view/vumls/home) below:

Speaker: Lincen Yang (https://www.lincen.nl)
Title: MDL-based Interpretable Machine Learning for High-stake Applications (please see the bio of Lincen and the abstract below the email)
Time: Monday, 27, Oct. 10:00
Room: NU-4A67

Lincen is visiting from Leiden, and he will stay around for the rest of the day. If you would like to have a chat with him, please let us know!

Looking forward to seeing all of you on Monday!

Best from the organizers,
Zhao & Daniel

——————
Bio: Lincen is a postdoctoral researcher at Leiden Institute of Advanced Computer Science (LIACS) and a guest researcher at Leiden University Medical Center (LUMC). His research lies at the intersection of interpretable machine learning and information-theoretic data mining. Specifically, he studies the Minimum Description Length (MDL) principle and develops human-centered algorithms that enable trustworthy knowledge discovery and high-stakes decision-making. Lincen also organized the inaugural Human-Centered Data Mining workshop at ECML-PKDD 2025.

Abstract: Applying machine learning to high-stakes domains such as healthcare and manufacturing requires models that are both interpretable and capable of quantifying uncertainty. This talk presents two related lines of research: (1) combining interpretability with uncertainty quantification through a tree-based conditional density estimation method, and (2) enhancing model comprehensibility via probabilistic rule sets that reduce the cognitive workload of understanding model behavior. Two real-world use cases—processor microarchitecture design in computer architecture and ICU readmission analysis in clinical informatics—illustrate their effectiveness in supporting data-driven decision-making. Built upon the Minimum Description Length (MDL) framework, these works leverage its hyperparameter-free formulation to enable interpretable, reliable, and easy-to-use methods for domain experts.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://listserver.vu.nl/pipermail/vu.ml.seminar.beta/attachments/20251024/4365aab4/attachment.html>