Improving novelty detection using the reconstructions of nearest neighbours
We show that using nearest neighbours in the latent space of autoencoders (AE) significantly improves performance of semi-supervised novelty detection in both single and multi-class contexts. Autoencoding methods detect novelty by learning to differentiate between the non-novel training class(es) an...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2022-07-01
|
| Series: | Array |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2590005622000388 |
| _version_ | 1811340755028934656 |
|---|---|
| author | Michael Mesarcik Elena Ranguelova Albert-Jan Boonstra Rob V. van Nieuwpoort |
| author_facet | Michael Mesarcik Elena Ranguelova Albert-Jan Boonstra Rob V. van Nieuwpoort |
| author_sort | Michael Mesarcik |
| collection | DOAJ |
| description | We show that using nearest neighbours in the latent space of autoencoders (AE) significantly improves performance of semi-supervised novelty detection in both single and multi-class contexts. Autoencoding methods detect novelty by learning to differentiate between the non-novel training class(es) and all other unseen classes. Our method harnesses a combination of the reconstructions of the nearest neighbours and the latent-neighbour distances of a given input’s latent representation. We demonstrate that our nearest-latent-neighbours (NLN) algorithm is memory and time efficient, does not require significant data augmentation, nor is reliant on pretrained networks. Furthermore, we show that the NLN-algorithm is easily applicable to multiple datasets without modification. Additionally, the proposed algorithm is agnostic to autoencoder architecture and reconstruction error method. We validate our method across several standard datasets for a variety of different autoencoding architectures such as vanilla, adversarial and variational autoencoders using either reconstruction, residual or feature consistent losses. The results show that the NLN algorithm grants up to a 17% increase in Area Under the Receiver Operating Characteristics (AUROC) curve performance for the multi-class case and 8% for single-class novelty detection. |
| first_indexed | 2024-04-13T18:46:37Z |
| format | Article |
| id | doaj.art-c38f33f22dde4311ad78276c5509e674 |
| institution | Directory Open Access Journal |
| issn | 2590-0056 |
| language | English |
| last_indexed | 2024-04-13T18:46:37Z |
| publishDate | 2022-07-01 |
| publisher | Elsevier |
| record_format | Article |
| series | Array |
| spelling | doaj.art-c38f33f22dde4311ad78276c5509e6742022-12-22T02:34:34ZengElsevierArray2590-00562022-07-0114100182Improving novelty detection using the reconstructions of nearest neighboursMichael Mesarcik0Elena Ranguelova1Albert-Jan Boonstra2Rob V. van Nieuwpoort3University of Amsterdam, Science Park 904, Amsterdam, 1098XH, The Netherlands; Corresponding author.eScience Center, Science Park 140, Amsterdam, 1098XG, The NetherlandsASTRON, the Netherlands Institute for Radio Astronomy, Oude Hoogeveensedijk 4, Dwingeloo, 7991PD, The NetherlandsUniversity of Amsterdam, Science Park 904, Amsterdam, 1098XH, The Netherlands; eScience Center, Science Park 140, Amsterdam, 1098XG, The NetherlandsWe show that using nearest neighbours in the latent space of autoencoders (AE) significantly improves performance of semi-supervised novelty detection in both single and multi-class contexts. Autoencoding methods detect novelty by learning to differentiate between the non-novel training class(es) and all other unseen classes. Our method harnesses a combination of the reconstructions of the nearest neighbours and the latent-neighbour distances of a given input’s latent representation. We demonstrate that our nearest-latent-neighbours (NLN) algorithm is memory and time efficient, does not require significant data augmentation, nor is reliant on pretrained networks. Furthermore, we show that the NLN-algorithm is easily applicable to multiple datasets without modification. Additionally, the proposed algorithm is agnostic to autoencoder architecture and reconstruction error method. We validate our method across several standard datasets for a variety of different autoencoding architectures such as vanilla, adversarial and variational autoencoders using either reconstruction, residual or feature consistent losses. The results show that the NLN algorithm grants up to a 17% increase in Area Under the Receiver Operating Characteristics (AUROC) curve performance for the multi-class case and 8% for single-class novelty detection.http://www.sciencedirect.com/science/article/pii/S2590005622000388Anomaly detectionAutoencodersSemi-supervised learning |
| spellingShingle | Michael Mesarcik Elena Ranguelova Albert-Jan Boonstra Rob V. van Nieuwpoort Improving novelty detection using the reconstructions of nearest neighbours Array Anomaly detection Autoencoders Semi-supervised learning |
| title | Improving novelty detection using the reconstructions of nearest neighbours |
| title_full | Improving novelty detection using the reconstructions of nearest neighbours |
| title_fullStr | Improving novelty detection using the reconstructions of nearest neighbours |
| title_full_unstemmed | Improving novelty detection using the reconstructions of nearest neighbours |
| title_short | Improving novelty detection using the reconstructions of nearest neighbours |
| title_sort | improving novelty detection using the reconstructions of nearest neighbours |
| topic | Anomaly detection Autoencoders Semi-supervised learning |
| url | http://www.sciencedirect.com/science/article/pii/S2590005622000388 |
| work_keys_str_mv | AT michaelmesarcik improvingnoveltydetectionusingthereconstructionsofnearestneighbours AT elenaranguelova improvingnoveltydetectionusingthereconstructionsofnearestneighbours AT albertjanboonstra improvingnoveltydetectionusingthereconstructionsofnearestneighbours AT robvvannieuwpoort improvingnoveltydetectionusingthereconstructionsofnearestneighbours |