Grégoire Montavon: Uncovering Input-Target Associations with Explainable AI

Speaker:
Grégoire Montavon (Charité / BIFOLD)

Title:
Uncovering Input-Target Associations with Explainable AI

Abstract:

Explainable AI (XAI) has become an essential technology for promoting transparency in machine learning (ML). However, Explainable AI offers more than just a framework for verifying ML models. It also provides a fitting framework for identifying input-target relationships in a broader range of systems of interest. Unlike basic correlational analyses, an XAI algorithm combined with a powerful ML model can filter out weak correlates in the data and identify combinations of input variables that are the most predictive. In this talk, I will present recent applications of this approach in the context of cancer to improve understanding of regulatory networks and treatment outcomes. In the second part of the presentation, I will transition from input-target relationships to input-uncertainty relationships (e.g. in electricity markets) and demonstrate how Explainable AI methods can be tailored for this type of analysis.

Read more about his research here.

Lecture Series:
SysBio Lecture Series: AI for Systems Medicine