解析焦点熵的数学特性，揭示其在不平衡数据中的强化机制
Functional Properties of the Focal-Entropy
- 从分布视角分析焦点熵的收敛性与凸性条件
- 证明最小值唯一存在且可能偏离真实数据分布
- 揭示其放大中等概率、抑制高概率的内在机制
焦点损失已广泛应用于计算机视觉中的类别不平衡分类问题。尽管其经验表现优异，但对其信息论特性的系统研究仍不完整。本文采用分布视角，研究焦点熵——交叉熵的焦点损失对应物。分析确立了焦点熵的有限性、凸性与连续性条件，并提供多种渐近刻画。证明了焦点熵最小值的存在性与唯一性，描述其结构特征，并表明其可显著偏离真实数据分布。特别地，我们严格证明：焦点损失会放大中等概率，抑制高概率结果；在极端类别不平衡下，进入过抑制阶段，使极小概率进一步被削弱。这些结果经实验验证，为理解焦点损失提供了理论基础，并厘清其在不平衡学习任务中引入的权衡关系。
原文摘要 · Abstract (English)
The focal-loss has become a widely used alternative to cross-entropy in class-imbalanced classification problems, particularly in computer vision. Despite its empirical success, a systematic information-theoretic study of the focal-loss remains incomplete. In this work, we adopt a distributional viewpoint and study the focal-entropy, a focal-loss analogue of the cross-entropy. Our analysis establishes conditions for finiteness, convexity, and continuity of the focal-entropy, and provides various asymptotic characterizations. We prove the existence and uniqueness of the focal-entropy minimizer, describe its structure, and show that it can depart significantly from the data distribution. In particular, we rigorously show that the focal-loss amplifies mid-range probabilities, suppresses high-probability outcomes, and, under extreme class imbalance, induces an over-suppression regime in which very small probabilities are further diminished. These results, which are also experimentally validated, offer a theoretical foundation for understanding the focal-loss and clarify the trade-offs that it introduces when applied to imbalanced learning tasks.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。