Wasserstein Distributionally Robust Performative Prediction
Siyi Wang, Zifan Wang, Karl H. Johansson · 2026
Read Wasserstein Distributionally Robust Performative Prediction by Siyi Wang, Zifan Wang, Karl H. Johansson (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-06730.txt.
Description
Siyi Wang, Zifan Wang, and Karl H. Johansson study predictions that change the behavior they predict. Once a model is deployed, people adapt to it, shifting the data distribution in ways that depend on the model itself. The paper builds a framework for this performative setting that stays robust when only limited data is available.
They place an uncertainty set around the observed data distribution, measured by Wasserstein distance, and train the model against the worst case inside it. The problem is recast as a tractable min-max task, and two iterative algorithms are given for reaching a stable point where retraining no longer moves the model. Both are shown to converge quickly under standard conditions, with bounds on how far the stable point sits from the true optimum. A credit scoring simulation illustrates the approach.
Catalog record
| CREATOR | Siyi Wang, Zifan Wang, Karl H. Johansson |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 75 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-06730