Causal Supervised Learning

Contact person: Johan Pensar
Keywords: Structural causal models; supervised machine learning; causal representation learning; out-of-domain generalization
Research group: Integreat
Department of Mathematics

There has lately been an increasing interest towards the integration of causal inference principles into the framework of supervised machine learning. While a machine learning model that is trained to minimize standard empirical risk typically performs well within its training domain, the performance can degrade considerably under various shifts in the domain (or environment). One of the key characteristics of causal relationships, as opposed to correlation-based relationships, is that they are thought to be stable or even invariant across domains. Thus, when a model is trained to capture and make its predictions based on causal relationships, it is more likely to generalize well to different contexts and scenarios outside its specific training domain. By drawing inspiration from the field of causality, the main goal of this research theme will be to develop new supervised learning methods that are robust in terms of how well they generalize to new domains.

Mentoring and internship will be offered by a relevant external partner.