论文标题
OOD - 探针：偏域概括的神经解释
OOD-Probe: A Neural Interpretation of Out-of-Domain Generalization
论文作者
论文摘要
概括不域外（OOD）的能力是深度神经网络发展的重要目标，研究人员提出了来自各种基础的许多高性能的OOD泛化方法。尽管许多OOD算法在各种情况下都表现良好，但这些系统被评估为``Black-Boxes''。取而代之的是，我们提出了一个灵活的框架，该框架使用探测模块来评估具有更精细粒度的OOD系统，该模块可预测中间表示的原始域。我们发现表示始终编码有关域的一些信息。虽然在不同的OOD算法上，层的编码模式在很大程度上保持稳定，但它们在数据集中有所不同。例如，有关旋转（在旋转的旋转）的信息是下层最明显的，而有关样式（在VLCS和PACS上）的信息在中间层上最为明显。此外，高探测结果与域的概括性能相关，从而导致了开发OOD泛化系统的进一步方向。
The ability to generalize out-of-domain (OOD) is an important goal for deep neural network development, and researchers have proposed many high-performing OOD generalization methods from various foundations. While many OOD algorithms perform well in various scenarios, these systems are evaluated as ``black-boxes''. Instead, we propose a flexible framework that evaluates OOD systems with finer granularity using a probing module that predicts the originating domain from intermediate representations. We find that representations always encode some information about the domain. While the layerwise encoding patterns remain largely stable across different OOD algorithms, they vary across the datasets. For example, the information about rotation (on RotatedMNIST) is the most visible on the lower layers, while the information about style (on VLCS and PACS) is the most visible on the middle layers. In addition, the high probing results correlate to the domain generalization performances, leading to further directions in developing OOD generalization systems.
