AdaProj: Adaptively Scaled Angular Margin Subspace Projections for Anomaly Detection with Auxiliary Classification Tasks
Keywords: representation learning, anomaly detection, semi-supervised learning, angular margin loss
Abstract: One of the state-of-the-art approaches for semi-supervised anomaly detection is to first learn an embedding space and then estimate the distribution of normal data. This can be done by using one-class losses or by using auxiliary classification tasks based on meta information or self-supervised learning. Angular margin losses are a popular training objective because they increase intra-class similarity and avoid learning trivial solutions by reducing inter-class similarity. In this work, AdaProj a novel loss function that generalizes upon angular margin losses is presented. In contrast to angular margin losses, which project data of each class as close as possible to their corresponding class centers, AdaProj learns to project data onto class-specific subspaces. By doing so, the resulting distributions of embeddings belonging to normal data are not required to be as restrictive as other loss functions allowing a more detailed view on the data. This enables a system to more accurately detect anomalous samples during testing. In experiments conducted on the DCASE2022 and DCASE2023 datasets, it is shown that using AdaProj to learn an embedding space significantly outperforms other commonly used loss functions achieving a new state-of-the-art performance on the DCASE2023 dataset.
Primary Area: representation learning for computer vision, audio, language, and other modalities
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 1177
Loading