← 研踪 PaperTrailarXiv:2611.12696 · v1
Generalized Adaptive Boosting and the Geometry of Mistakes
首次提交 2026-11-10 00:56
版本更新 2026-11-10 00:56
雷达观测 2026-11-10 15:22
摘要
Not all mistakes are alike. For example, in medical diagnosis, false positive and false negative mistakes have a different impact and should not be aggregated. By revisiting the classical theory of boosting through this lens, we introduce the notion of -learner, which returns weak hypotheses achieving joint false-positive/false-negative guarantees from some set . This formulation generalizes the standard notion of -weak learning--recovered by taking --and allows for more flexible notions of weak learnability that account for the impact of the distribution on the achievable learning guarantees. This raises a natural question: which sets are boostable? We answer this by giving a complete geometric characterization that unifies, as special cases, both the classical boosting dichotomy and the recent multi-objective extension of Bressan et al. (COLT 2025). We also present an adaptive algorithm for boosting -learners without relying on the prior knowledge of , generalizing AdaBoost's ability to handle unknown margin parameters. Finally, we demonstrate that -learners can achieve confidence amplification. However, the analysis is significantly more involved than in classical boosting and relies on the centerpoint theorem from discrete and computational geometry.
作者备注:Accepted at NeurIPS 2026
观测到的版本
相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。
v1 · 2026-11-10 00:56 · 最新观测版本
Generalized Adaptive Boosting and the Geometry of Mistakes
Not all mistakes are alike. For example, in medical diagnosis, false positive and false negative mistakes have a different impact and should not be aggregated. By revisiting the classical theory of boosting through this lens, we introduce the notion of -learner, which returns weak hypotheses achieving joint false-positive/false-negative guarantees from some set . This formulation generalizes the standard notion of -weak learning--recovered by taking --and allows for more flexible notions of weak learnability that account for the impact of the distribution on the achievable learning guarantees. This raises a natural question: which sets are boostable? We answer this by giving a complete geometric characterization that unifies, as special cases, both the classical boosting dichotomy and the recent multi-objective extension of Bressan et al. (COLT 2025). We also present an adaptive algorithm for boosting -learners without relying on the prior knowledge of , generalizing AdaBoost's ability to handle unknown margin parameters. Finally, we demonstrate that -learners can achieve confidence amplification. However, the analysis is significantly more involved than in classical boosting and relies on the centerpoint theorem from discrete and computational geometry.
查看此版本解读 →查看此版本原文 ↗