用向量量化构建分区域校准，提升多分类模型准确性。
Divide et Calibra: Multiclass Local Calibration via Vector Quantization
- 通过向量量化划分表示空间，共享编码因子实现区域校准
- 在基准数据集上局部校准显著提升，稀疏区域也表现良好
- 适合对校准精度要求高的医疗、金融等高风险场景
高风险场景中，机器学习模型的准确性和校准性至关重要，但多分类校准仍具挑战：全局方法假设校准误差在潜在空间中均匀分布，而局部方法常依赖降维，导致信息丢失。为此，我们提出一种组合式多分类校准方法，通过共享的码字相关因子构建区域特异性校准映射。我们基于向量量化（VQ）实现表示空间的结构化划分，并采用索引参数化狄利克雷浓度，实现区域间的参数共享。该方法学习到具有异质性的校准映射，在稀疏潜在空间区域仍具有良好泛化能力。在基准数据集上的实验表明，局部校准性能显著提升，同时保持了竞争性的全局校准与预测表现。
原文摘要 · Abstract (English)
Accurate and well-calibrated Machine Learning (ML) models are mandatory in high-stakes settings, yet effective multiclass calibration remains challenging: global approaches assume calibration errors are homogeneous across the latent space, while local methods often rely on latent-space dimensionality reduction, which leads to information loss. To address these issues, we propose a compositional approach to multiclass calibration, where region-specific calibration maps are constructed from shared codeword-dependent factors. We instantiate this idea via Vector Quantization (VQ), which induces a structured partition of the representation space, and an indexed parameterization of Dirichlet concentrations that enables parameter sharing across regions. Our approach learns heterogeneous calibration maps that generalize well even to sparse regions of the latent space. Experiments on benchmark datasets show significant improvements in local calibration while maintaining competitive global calibration and predictive performance.
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。