Deconfounding and Causal Regularization for Stability and External Validity
This work tackles problems of stability and external validity in machine learning for researchers and practitioners dealing with heterogeneous data, but it is incremental as it reviews and synthesizes existing techniques.
The paper reviews recent work on removing hidden confounding and causal regularization to improve stability, replicability, and distributional robustness in heterogeneous data, addressing issues like concept drift when data distributions change.
We review some recent work on removing hidden confounding and causal regularization from a unified viewpoint. We describe how simple and user-friendly techniques improve stability, replicability and distributional robustness in heterogeneous data. In this sense, we provide additional thoughts to the issue on concept drift, raised by Efron (2020), when the data generating distribution is changing.