Novelty Detection via Contrastive Learning with Negative Data Augmentation

Chen, Chengwei; Xie, Yuan; Lin, Shaohui; Qiao, Ruizhi; Zhou, Jian; Tan, Xin; Zhang, Yi; Ma, Lizhuang

doi:10.24963/ijcai.2021/84

Novelty Detection via Contrastive Learning with Negative Data Augmentation

Chengwei Chen, Yuan Xie, Shaohui Lin, Ruizhi Qiao, Jian Zhou, Xin Tan, Yi Zhang, Lizhuang Ma

Download BibTeX Cite

Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence. Main Track. Pages 606-614.

https://doi.org/10.24963/ijcai.2021/84

Keywords

Computer Vision: 2D and 3D Computer Vision
Data Mining: Anomaly/Outlier Detection

Abstract

Novelty detection is the process of determining whether a query example differs from the learned training distribution. Previous generative adversarial networks based methods and self-supervised approaches suffer from instability training, mode dropping, and low discriminative ability. We overcome such problems by introducing a novel decoder-encoder framework. Firstly, a generative network (decoder) learns the representation by mapping the initialized latent vector to an image. In particular, this vector is initialized by considering the entire distribution of training data to avoid the problem of mode-dropping. Secondly, a contrastive network (encoder) aims to ``learn to compare'' through mutual information estimation, which directly helps the generative network to obtain a more discriminative representation by using a negative data augmentation strategy. Extensive experiments show that our model has significant superiority over cutting-edge novelty detectors and achieves new state-of-the-art results on various novelty detection benchmarks, e.g. CIFAR10 and DCASE. Moreover, our model is more stable for training in a non-adversarial manner, compared to other adversarial based novelty detection methods.

Chengwei Chen, Yuan Xie, Shaohui Lin, Ruizhi Qiao, Jian Zhou, Xin Tan, Yi Zhang, and Lizhuang Ma. 2021. Novelty Detection via Contrastive Learning with Negative Data Augmentation. In Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence, IJCAI-21. IJCAI, 606–614. https://doi.org/10.24963/ijcai.2021/84

@inproceedings{ijcai2021p84,
  title     = {Novelty Detection via Contrastive Learning with Negative Data Augmentation},
  author    = {Chen, Chengwei and Xie, Yuan and Lin, Shaohui and Qiao, Ruizhi and Zhou, Jian and Tan, Xin and Zhang, Yi and Ma, Lizhuang},
  booktitle = {Proceedings of the Thirtieth International Joint Conference on
               Artificial Intelligence, {IJCAI-21}},
  publisher = {International Joint Conferences on Artificial Intelligence Organization},
  editor    = {Prof. Zhi-Hua Zhou},
  pages     = {606--614},
  year      = {2021},
  month     = {8},
  note      = {Main Track},
  doi       = {10.24963/ijcai.2021/84},
  url       = {https://doi.org/10.24963/ijcai.2021/84},
}

TY  - CPAPER
TI  - Novelty Detection via Contrastive Learning with Negative Data Augmentation
AU  - Chen, Chengwei
AU  - Xie, Yuan
AU  - Lin, Shaohui
AU  - Qiao, Ruizhi
AU  - Zhou, Jian
AU  - Tan, Xin
AU  - Zhang, Yi
AU  - Ma, Lizhuang
T2  - Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence, IJCAI-21
PB  - International Joint Conferences on Artificial Intelligence Organization
PY  - 2021
SP  - 606
EP  - 614
DO  - 10.24963/ijcai.2021/84
UR  - https://doi.org/10.24963/ijcai.2021/84
ER  -