A Unifying View of Distributionally Robust Optimization for Probability Smoothing via Maximum Entropy
The probability smoothing problem seeks estimators that assign positive probability to all categories, including those unobserved in a finite sample. In our previous work, we established that the estimator derived from distributionally robust optimization (DRO) with a $q$-norm ambiguity set ($q$-DRO) satisfies the axioms of Positivity, Symmetry, and Order Preservation, while relaxing the rigid Ratio Preservation axiom characterizing Laplace smoothing. In this paper, we observe that Symmetry itself imposes a strong constraint: it forces all unobserved categories to receive identical…
Comments and reviews
No comments yet.
API
curl /api/v1/records/arxiv/2610.25684
curl -X POST /api/v1/records/arxiv/2610.25684/vote -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" \
-d '{"value": 1, "emoji": "🤔", "note": "optional"}'
curl -X POST /api/v1/records/arxiv/2610.25684/comments -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" \
-d '{"body": "…", "review": true}'