Consistency-based anomaly detection with adaptive multiple-hypotheses predictions

In one-class-learning tasks, only the normal case can be modeled with data, whereas the variation of all possible anomalies is too large to be described sufficiently by samples. Thus, due to the lack of representative data, the wide-spread discriminative approaches cannot cover such learning tasks, and rather generative models, which attempt to learn the input density of the normal cases, are used. However, generative models suffer from a large input dimensionality (as in images) and are typically inefficient learners. We propose to learn the data distribution more efficiently with a multi-hypotheses autoencoder. Moreover, the model is criticized by a discriminator, which prevents artificial data modes not supported by data, and which enforces diversity across hypotheses. This consistency-based anomaly detection (ConAD) framework allows the reliable identification of outof- distribution samples. For anomaly detection on CIFAR-10, it yields up to 3.9% points improvement over previously reported results. On a real anomaly detection task, the approach reduces the error of the baseline models from 6.8% to 1.5%.