信号路由温度缩放：低容量风险条件校准用于小验证预算
Signal-Routed Temperature Scaling: Low-Capacity Risk-Conditioned Calibration for Small Validation Budgets
浏览论文内容
中文总结 AI 辅助
针对小验证预算下事后校准器容量选择问题，提出10参数信号路由温度缩放（SRTS-BCE），在CIFAR-100等基准上以低容量匹配高容量方法性能，并揭示容量偏好随数据量变化。
中文摘要 AI 辅助
当分类器仅从几千个留出样本进行重新校准时，校准映射的容量成为一个统计设计选择，而非纯粹的结构性选择：标量映射可能欠拟合结构化的残差误校准，而高度自适应的映射则难以从如此小的数据划分中可靠估计。我们将校准目标与自适应容量分离，提出信号路由温度缩放（SRTS-BCE），这是一种10参数、保持argmax的校准器，它基于六个logit统计量对正确性风险评分进行交叉拟合，并为每个K=3风险组拟合一个顶层标签BCE温度，将TvA-TS恢复为其K=1极限。在微调的CIFAR-100 / ViT-B/16上，SRTS-BCE将ECE_15从1.65（标量TvA-TS）降至0.96，在全校准预算下匹配更高容量的SMART+BCE头（0.95）。随着预算缩小，两种机制分离：在n=250时，SRTS-BCE在所有三个CIFAR-100骨干网络上优于SMART+BCE（种子到抽样的分层区间排除零），而与标量的旗舰比较仍具方向性。一项协议冻结的Tiny-ImageNet后续实验重现了小预算分离，并在Swin-T上表现出依赖预算的排名反转；匹配的路由和映射控制表明，该效应既与学习路由器无关，也与离散分组无关。综合来看，这些结果将事后校准器容量识别为有限样本设计选择，其首选水平随可用校准数据量而变化。
英文摘要
When a classifier is recalibrated from only a few thousand held-out examples, the capacity of the calibration map becomes a statistical design choice rather than a purely architectural one: a scalar map can underfit structured residual miscalibration, while a highly adaptive map can be hard to estimate reliably from so small a split. We disentangle the calibration objective from adaptive capacity and propose signal-routed temperature scaling (SRTS-BCE), a 10-parameter, argmax-preserving calibrator that cross-fits a correctness-risk score over six logit statistics and fits one top-label-BCE temperature per $K=3$ risk groups, recovering TvA-TS as its $K=1$ limit. On fine-tuned CIFAR-100 / ViT-B/16, SRTS-BCE reduces $\mathrm{ECE}_{15}$ from 1.65 (scalar TvA-TS) to 0.96, matching the higher-capacity SMART+BCE head (0.95) at the full calibration budget. The two regimes separate as the budget shrinks: at $n=250$ SRTS-BCE beats SMART+BCE on all three CIFAR-100 backbones (the seed-to-draw hierarchical interval excludes zero), whereas the flagship comparison against the scalar remains directional. A protocol-frozen Tiny-ImageNet follow-up reproduces the small-budget separation and exhibits a budget-dependent ranking reversal on Swin-T; matched routing and map controls show that the effect is tied neither to the learned router nor to discrete grouping. Together the results identify post-hoc calibrator capacity as a finite-sample design choice whose preferred level shifts with the amount of available calibration data.
发表机构
- Adelaide University（阿德莱德大学）
机构由 AI 辅助整理，请以论文原文为准。
