Distributional robustness in inference
2019-07-12 — 2022-08-30
Wherein distributional mis-specification within a Wasserstein ball is treated, and ties to causal inference, differential privacy, and adversarial learning are recorded.
Placeholder, for inference which is robust to a mis-specification of distribution up to some ball in some probability metric. I saw Jose Blanchet present on Wasserstein distributional robustness at MCM 2019. Bookmarked for later.
To understand: Connections to causal inference, differential privacy, adversarial learning, hedging in portfolios…
Also to understand: How old is distributional robustness? I feel like when I discovered this literature it was already old, but no one seems to agree upon a specific foundational publication.