Out-of-Distribution Detection Using Union of 1-Dimensional Subspaces
Out-of-Distribution Detection Using Union of 1-Dimensional Subspaces
复制标题
DOI：
10.1109/cvpr46437.2021.00933
复制
发表时间：
2021-06
期刊：
影响因子：
--
通讯作者：
Alireza Zaeemzadeh;N. Bisagno;Zeno Sambugaro;N. Conci;Nazanin Rahnavard;M. Shah
中科院分区：
文献类型：
--
作者：
Alireza Zaeemzadeh;N. Bisagno;Zeno Sambugaro;N. Conci;Nazanin Rahnavard;M. Shah
分布外（OOD）检测的目标是处理测试样本来自与训练数据不同分布的情况。在本文中，我们认为如果将训练数据嵌入到低维空间中，使得嵌入的训练样本位于一维子空间的并集上，那么OOD样本就能更容易被检测到。我们表明这种对分布内（ID）样本的嵌入为我们提供了两个主要优势。首先，由于在特征空间中的紧凑表示，OOD样本不太可能占据与已知类别相同的区域。其次，属于一维子空间的ID样本的第一奇异向量可以用作它们的稳健代表。受这些观察结果的启发，我们训练一个深度神经网络，使得ID样本被嵌入到一维子空间的并集上。在测试时，利用深度学习中用于近似贝叶斯推理的采样技术，如果输入样本以概率0占据对应于ID样本的区域，则将其检测为OOD。ID样本的频谱成分被用作该区域的稳健代表。我们的方法没有任何需要使用额外信息调整的超参数，并且它可以在不同模态上应用，只需进行最小的改动。所提出方法的有效性在图像和视频分类领域的不同基准数据集上得到了证明。
The goal of out-of-distribution (OOD) detection is to handle the situations where the test samples are drawn from a different distribution than the training data. In this paper, we argue that OOD samples can be detected more easily if the training data is embedded into a low-dimensional space, such that the embedded training samples lie on a union of 1-dimensional subspaces. We show that such embedding of the in-distribution (ID) samples provides us with two main advantages. First, due to compact representation in the feature space, OOD samples are less likely to occupy the same region as the known classes. Second, the first singular vector of ID samples belonging to a 1-dimensional subspace can be used as their robust representative. Motivated by these observations, we train a deep neural network such that the ID samples are embedded onto a union of 1-dimensional subspaces. At the test time, employing sampling techniques used for approximate Bayesian inference in deep learning, input samples are detected as OOD if they occupy the region corresponding to the ID samples with probability 0. Spectral components of the ID samples are used as robust representative of this region. Our method does not have any hyperparameter to be tuned using extra information and it can be applied on different modalities with minimal change. The effectiveness of the proposed method is demonstrated on different benchmark datasets, both in the image and video classification domains.