Machine learning models should be explainable and robust on out-of-distribution samples, especially on safety-critical tasks such as healthcare, and security. However, current models heavily rely on i.i.d assumption, and are therefore sensitive to OOD data. In this talk, Wei Chen, from the Computing and Learning Theory group at Microsoft Research Asia, will show how causal inference tools can be leveraged to empower machine learning models and make them more robust. To achieve this goal, we propose the causal invariance model, which can eliminate spurious correlations and keep only causal relation for prediction, and we will show both theoretical and empirical proof.
Learn more about the 2021 Microsoft Research Summit: https://Aka.ms/researchsummit