CLUE：通过学习不确定性-误差对齐进行神经网络校准
CLUE: Neural Networks Calibration via Learning Uncertainty-Error alignment
AI总结：
提出CLUE方法，通过在训练中显式对齐预测不确定性与观测误差并联合优化预测与校准，实现了跨领域、低开销且校准质量优越的神经网络不确定性估计。
摘要：
可靠的不确定性估计对于在现实应用中部署神经网络（NN）至关重要。现有的校准技术通常依赖于事后调整或粗粒度分箱方法，在跨领域的可扩展性、可微性和泛化能力上仍然受限。在本研究中，我们提出了CLUE（Calibration via Learning Uncertainty-Error Alignment，通过学习不确定性-误差对齐进行校准），这是一种新颖的方法，在训练期间显式地将预测不确定性与观测误差对齐，其基于一个原则：校准良好的模型应产生与其经验损失相匹配的不确定性估计。CLUE采用了一种新颖的损失函数，使用不确定性和损失的汇总统计量作为代理，联合优化预测性能和校准。所提出的方法是完全可微的、领域无关的，并且与标准训练流程兼容。通过对视觉、回归和语言建模任务进行广泛实验，包括分布外和领域偏移场景，我们证明CLUE实现了卓越的校准质量和具有竞争力的预测性能，且相对于最先进的方法没有施加显著的计算开销。
英文摘要：
Reliable uncertainty estimation is critical for deploying neural networks (NNs) in real-world applications. While existing calibration techniques often rely on post-hoc adjustments or coarse-grained binning methods, they remain limited in scalability, differentiability, and generalization across domains. In this work, we introduce CLUE (Calibration via Learning Uncertainty-Error Alignment), a novel approach that explicitly aligns predicted uncertainty with observed error during training, grounded in the principle that well-calibrated models should produce uncertainty estimates that match their empirical loss. CLUE adopts a novel loss function that jointly optimizes predictive performance and calibration, using summary statistics of uncertainty and loss as proxies. The proposed method is fully differentiable, domain-agnostic, and compatible with standard training pipelines. Through extensive experiments on vision, regression, and language modeling tasks, including out-of-distribution and domain-shift scenarios, we demonstrate that CLUE achieves superior calibration quality and competitive predictive performance with respect to state-of-the-art approaches without imposing significant computational overhead.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
