📄 中文摘要
不确定性量化在模型选择、正则化以及量化预测不确定性以进行主动学习或OOD检测等任务中起着关键作用。在考虑将不确定性建模为概率集合的信念方法中，上熵作为一种不确定性度量发挥着核心作用。研究提供了上熵的计算方面的深入分析，给出了问题的全面算法和复杂性分析。特别地，证明了该问题具有强多项式解，并提出了相较于过去针对2-单调下概率及其特定情况的算法的多项显著改进。
2-单调下概率的上熵
出处: Upper Entropy for 2-Monotone Lower Probabilities
发布: 2026年3月26日
📄 English Summary
Upper Entropy for 2-Monotone Lower Probabilities
Uncertainty quantification is crucial for tasks such as model selection, regularization, and quantifying prediction uncertainties for active learning or out-of-distribution (OOD) detection. Within credal approaches that model uncertainty as sets of probabilities, upper entropy serves as a central uncertainty measure. This research provides a comprehensive algorithmic and complexity analysis of upper entropies, demonstrating that the problem has a strongly polynomial solution. Significant improvements over previous algorithms for 2-monotone lower probabilities and their specific cases are proposed.