This was part of New Horizons on Model Transportability and Data Integration

Learning Surrogate Indices from Historical A/Bs: Adversarial ML for Debiased Inference on Functionals of Ill-Posed Inverses

Nathan Kallus, Cornell University