Latent-Insensitive autoencoders for Anomaly Detection
This work addresses anomaly detection for domains with similar unlabeled data, presenting an incremental improvement over existing reconstruction-based methods.
The paper tackles the problem of anomaly detection in complex datasets with high inter-class variance by introducing Latent-Insensitive autoencoders (LIS-AE), which use unlabeled data from similar domains as negative examples to shape the latent layer, resulting in significant performance improvements as shown through quantitative and qualitative analysis.
Reconstruction-based approaches to anomaly detection tend to fall short when applied to complex datasets with target classes that possess high inter-class variance. Similar to the idea of self-taught learning used in transfer learning, many domains are rich with similar unlabelled datasets that could be leveraged as a proxy for out-of-distribution samples. In this paper we introduce Latent-Insensitive autoencoder (LIS-AE) where unlabeled data from a similar domain is utilized as negative examples to shape the latent layer (bottleneck) of a regular autoencoder such that it is only capable of reconstructing one task. We provide theoretical justification for the proposed training process and loss functions along with an extensive ablation study highlighting important aspects of our model. We test our model in multiple anomaly detection settings presenting quantitative and qualitative analysis showcasing the significant performance improvement of our model for anomaly detection tasks.