CLEX：面向噪声面部表情识别的互补标签交换学习
CLEX: Complementary Label Exchange Learning for Noisy Facial Expression Recognition
针对野外表情识别中的标签噪声问题，提出通过交换增强视图间的非目标logit来构造一致性约束，并配合尺度不变归一化和互补抑制损失，为视觉分类任务中利用非目标预测抑制虚假激活提供了可复用的正则化手段。
原文依据 Stochastic Non-Target Logit Exchange randomly swaps a subset of non-target logits between original and augmented views
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论野外面部表情识别（FER）严重受限于标签噪声和标注歧义。现有方法，包括样本选择、标签集成和一致性正则化，主要依赖普通标签监督，对非目标预测的控制有限，导致虚假激活和对噪声标签的过拟合。为解决这一局限，我们提出了一种新的学习框架，名为互补标签交换学习（CLEX），通过交换增强视图间非目标预测的知识来增强鲁棒性。具体而言，CLEX包含三个协同组件。首先，随机非目标Logit交换在原始视图和增强视图之间随机交换一部分非目标logit，以耦合易出错的预测，创建鲁棒的一致性约束。其次，尺度不变Logit归一化通过Lp范数归一化消除幅度伪影，确保正则化在几何有意义的方向上操作，而不是被任意尺度主导。第三，互补抑制损失在随机保留的非目标类子集上选择性惩罚虚假激活，避免妨碍判别学习的均匀收缩。为进一步稳定训练，我们加入了注意力一致性正则化，强制增强视图之间的空间对齐，同时保留辅助交叉熵以保持语义定位能力。在多个基准FER数据集（RAF-DB、FERPlus和AffectNet）上的大量实验表明，CLEX始终优于现有的鲁棒FER学习方法。
英文原文摘要 1695 字符
Facial expression recognition (FER) in the wild is severely hampered by label noise and annotation ambiguity. Existing methods, including sample selection, label ensembling, and consistency regularization, primarily rely on ordinary label supervision and offer limited control over non-target predictions, leading to spurious activations and overfitting to noisy labels. To address this limitation, we propose a novel learning framework, named Complementary Label Exchange Learning (CLEX), to enhance robustness by exchanging knowledge from non-target predictions across augmented views. Specifically, CLEX comprises three synergistic components. First, Stochastic Non-Target Logit Exchange randomly swaps a subset of non-target logits between original and augmented views to couple error-prone predictions, creating robust consistency constraints. Second, Scale-Invariant Logit Normalization eliminates magnitude artifacts through Lp-norm normalization, ensuring that regularization operates over geometrically meaningful directions rather than being dominated by arbitrary scales. Third, Complementary Suppression Loss selectively penalizes spurious activations over a randomly retained subset of non-target classes, avoiding the uniform shrinkage that hampers discriminative learning. To further stabilize training, we incorporate attention consistency regularization that enforces spatial alignment between augmented views, while retaining auxiliary cross-entropy to preserve semantic localization capability. Extensive experiments across multiple benchmark FER datasets (RAF-DB, FERPlus, and AffectNet) demonstrate that CLEX consistently outperforms existing robust FER learning approaches.