用扰动+校正生成多样预测，提升模型在分布外时的鲁棒性。
Perturb and Correct: Post-Hoc Ensembles using Affine Redundancy
- 对隐藏层随机扰动后，用最小二乘法校正输出层，实现预测多样性。
- 在分布内数据上保持一致，在分布外差异显著，实现良好内外部平衡。
- 仅需一个预训练模型，适合部署受限场景下的可靠性增强。
在分布内数据上表现相似的模型，在分布外可能行为迥异。我们提出扰动与校正（P&C）方法，从单一预训练网络构建认知上多样的预测器。P&C 在隐藏层施加随机扰动，并在后续仿射层进行最小二乘校正，使预测器在校准数据上达成一致，而在分布外自由分歧。通过分析校正后的残差及其一阶敏感度发现：残差在分布附近受杠杆项控制，而修正后的敏感度随输入偏离校准几何而增长。实验表明，P&C 在 MuJoCo 动力学预测和 CIFAR-10 分布外检测任务中均实现优异的分布内/外权衡，性能达到或超越标准后处理基线，且仅需单个预训练模型。结果揭示了利用过参数化作为深度学习优势的潜力。
原文摘要 · Abstract (English)
Models that are indistinguishable on in-distribution data can behave very differently under distribution shift. We introduce Perturb-and-Correct (P&C), a post-hoc method for constructing epistemically diverse predictors from a single pretrained network. P&C applies random hidden layer perturbations with a least-squares correction in the subsequent affine layer, producing predictors that agree on calibration data while remaining free to disagree away from it. We analyze this mechanism through the post-correction residual and its first-order sensitivity: the residual is controlled near the calibration distribution by a leverage term, while corrected sensitivity grows as inputs deviate from the calibration geometry. Empirically, P&C achieves a strong ID/OOD tradeoff across MuJoCo dynamics prediction and CIFAR-10 OOD detection, matching or outperforming standard post-hoc baselines while requiring only a single pretrained model. Our findings highlight the potential in further exploiting overparameterization as a strength of deep learning models.
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。