并非所有故障模式都是平等的：训练深度神经网络进行可解释（Mis）分类
Not all Failure Modes are Created Equal: Training Deep Neural Networks for Explicable (Mis)Classification
摘要：
深度神经网络在图像分类任务中通常很脆弱，并且已知会对输入进行错误分类。虽然这些错误分类可能是不可避免的，但不能认为所有故障模式都是相同的。一些错误的分类（例如，将狗的图像分类为飞机）会使人困惑，并导致人对系统失去信任。更糟糕的是，这些错误（例如，一个人被错误地归类为灵长类动物）会产生令人厌恶的社会影响。因此，在这项工作中，我们的目标是减少无法解释的错误。为了解决这一挑战，我们首先讨论获得类级语义的方法，该方法捕获人类关于哪些类在语义上接近{ \em vs.}的期望（$M^h$）。那些遥远的地方我们表明，对于流行的图像基准测试（如CIFAR-10，CIFAR-100，ImageNet），类级语义可以很容易地通过利用人类主题研究或公开可用的人类策划的知识库来获得。其次，我们提出了使用加权损失函数（WLF）来惩罚错误分类的权重，他们的费解。最后，我们表明，使用所提出的方法训练（或微调）现有的分类器会导致深度神经网络具有（1）相当的前1精度，（2）在分布和分布外（OOD）测试数据上更容易解释的故障模式，以及（3）与现有工作相比，在收集额外的人类标签方面产生的成本显着降低。
英文摘要：
Deep Neural Networks are often brittle on image classification tasks and known to misclassify inputs. While these misclassifications may be inevitable, all failure modes cannot be considered equal. Certain misclassifications (eg. classifying the image of a dog to an airplane) can perplex humans and result in the loss of human trust in the system. Even worse, these errors (eg. a person misclassified as a primate) can have odious societal impacts. Thus, in this work, we aim to reduce inexplicable errors. To address this challenge, we first discuss methods to obtain the class-level semantics that capture the human's expectation ($M^h$) regarding which classes are semantically close {\em vs.} ones that are far away. We show that for popular image benchmarks (like CIFAR-10, CIFAR-100, ImageNet), class-level semantics can be readily obtained by leveraging either human subject studies or publicly available human-curated knowledge bases. Second, we propose the use of Weighted Loss Functions (WLFs) to penalize misclassifications by the weight of their inexplicability. Finally, we show that training (or fine-tuning) existing classifiers with the proposed methods lead to Deep Neural Networks that have (1) comparable top-1 accuracy, (2) more explicable failure modes on both in-distribution and out-of-distribution (OOD) test data, and (3) incur significantly less cost in the gathering of additional human labels compared to existing works.
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。
