ELITE: Robust Deep Anomaly Detection with Meta Gradient
ELITE: Robust Deep Anomaly Detection with Meta Gradient
复制标题
ELITE：使用元梯度进行稳健的深度异常检测
DOI：
10.1145/3447548.3467320
复制
发表时间：
2021
期刊：
影响因子：
--
通讯作者：
Rundensteiner, Elke A.
中科院分区：
文献类型：
--
作者：
Zhang, Huayi;Cao, Lei;VanNostrand, Peter;Madden, Samuel;Rundensteiner, Elke A.
深度学习技术已被广泛用于从复杂数据中检测异常。这些技术中的大多数是无监督或半监督的，因为缺乏大量的标记异常。然而，它们通常依赖于未被异常污染的干净训练数据来学习正常数据的分布。否则，学习的分布往往会失真，因此无法有效区分正常和异常数据。为了解决这个问题，我们提出了一种称为ELITE的新方法，该方法使用少量标记的示例来推断隐藏在训练样本中的异常。然后，它将这些异常转化为有用的信号，帮助更好地从用户数据中检测异常。与经典的半监督分类策略不同，ELITE使用标记的样本作为训练数据，将其作为验证集。它利用验证损失的梯度来预测一个训练样本是否异常。直觉是，正确识别隐藏的异常可以产生更好的深度异常模型，减少验证损失。我们在公共基准数据集上的实验表明，与最先进的方法相比，ELITE在ROC AUC方面实现了高达30%的改进，但对污染的训练数据具有鲁棒性。
Deep Learning techniques have been widely used in detecting anomalies from complex data. Most of these techniques are either unsupervised or semi-supervised because of a lack of a large number of labeled anomalies. However, they typically rely on a clean training data not polluted by anomalies to learn the distribution of the normal data. Otherwise, the learned distribution tends to be distorted and hence ineffective in distinguishing between normal and abnormal data. To solve this problem, we propose a novel approach called ELITE that uses a small number of labeled examples to infer the anomalies hidden in the training samples. It then turns these anomalies into useful signals that help to better detect anomalies from user data. Unlike the classical semi-supervised classification strategy which uses labeled examples as training data, ELITE uses them as validation set. It leverages the gradient of the validation loss to predict if one training sample is abnormal. The intuition is that correctly identifying the hidden anomalies could produce a better deep anomaly model with reduced validation loss. Our experiments on public benchmark datasets show that ELITE achieves up to 30% improvement in ROC AUC comparing to the state-of-the-art, yet robust to polluted training data.