On Fisher Consistency of Surrogate Losses for Optimal Dynamic Treatment Regimes with Multiple Categorical Treatments per Stage
Nilanjana Laha ??,??, Nilson Chapagain ??, Victoria Cicherski ??, Aaron Sonabend-W ?? · 2025
Read On Fisher Consistency of Surrogate Losses for Optimal Dynamic Treatment Regimes with Multiple Categorical Treatments per Stage by Nilanjana Laha ??,??, Nilson Chapagain ??, Victoria Cicherski ??, Aaron Sonabend-W ?? (2025) 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-2505-17285.txt.
Introduction
Patients with chronic illness are treated in stages, and learning the best sequence of choices from past records is the dynamic treatment regime problem. With many stages and many treatment options per stage, the task becomes a chain of weighted classification problems. The authors ask which substitute loss functions can stand in for the true classification goal without bending the answer, a property called Fisher consistency. They work within stagewise separable, non-negative losses and find exact conditions for the property to hold.
Many popular convex losses fail the test, including smooth and relative-margin families, and the paper proves the failures formally. As a constructive answer, the authors propose a direct search method built on smooth, non-convex surrogates, with a gradient algorithm and a sharp bound on its regret decay. Simulations and a sepsis fluid study from hospital records show the method at work.
Catalog record
| CREATOR | Nilanjana Laha ??,??, Nilson Chapagain ??, Victoria Cicherski ??, Aaron Sonabend-W ?? |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 666 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-17285