RESEARCHING LIKELIHOOD RATIOS METHOD AND DEEP LEARNING  FOR OUT-OF-DISTRIBUTION DETECTION

Do Thi Thu; Dao Minh Tuan; Dang Viet Hung

Do Thi Thu Hung Yen University of Technology and Education
Dao Minh Tuan Hung Yen University of Technology and Education
Dang Viet Hung Hung Yen University of Technology and Education

Keywords: Likelihood Ratios, Anomaly Data, Anomaly Distribution, Anomaly Data Detection, Machine Learning Models

Abstract

This paper focuses on the integration of two important methods: Likelihood Ratios and deep learning, to address the problem of detecting abnormal distributions in image data using the CIFAR-10 and CIFAR-100 datasets. The introduction provides an overview of the issue and the necessity of combining these methods, emphasizing their comprehensive and integrated nature in tackling complex problems. Subsequently, the paper details the use of Likelihood Ratios to estimate event probabilities, coupled with deep learning models such as GMM to detect abnormal distributions in the data, highlighting the accuracy and flexibility of the approach. The following section describes the process of selecting suitable models, as well as optimization through the integration of the two aforementioned methods. The performance evaluation of the proposed model is conducted using metrics such as Confusion Matrix, ROC, AUROC, AUPR on CIFAR-10 and
CIFAR-100, demonstrating a deep and comprehensive analysis of the proposed method’s performance. Finally, the paper draws overall conclusions and proposes potential directions for further development, opening up numerous opportunities for application and continued research in this field. This represents a significant step towards shaping an effective and in-depth approach to addressing the complex problem of abnormal distribution detection in image data.

References

Ren, J., Liu, P.J., Fertig, E., Snoek, J., Poplin, R., DePristo, M.A., Dillon, J.V., & Lakshminarayanan, B., Likelihood Ratios for Out-of-Distribution Detection. ArXiv, abs/1906.02845, 2019.

Reynolds, D.A., A Gaussian Mixture Modeling Approach to Text-Independent Speaker Identification. PhD thesis, Georgia Institute of Technology, 1992.

Petra Bevandi´c, Ivan Krešo, Marin Orši´c, and Siniša Šegvi´c. Discriminative out-of-distribution detection for semantic segmentation. arXiv preprint arXiv:1808.07703, 2018.

Jiefeng Chen, Yixuan Li, Xi Wu, Yingyu Liang, and Somesh Jha. Informative outlier matters: Robustifying out-of-distribution detection using outlier mining. arXiv preprint arXiv:2006.15207, 2020.

Dan Hendrycks and Kevin Gimpel. A baseline for detecting misclassified and out-of-distribution examples in neural networks. arXiv preprint arXiv:1610.02136, 2016.

Dan Hendrycks, Mantas Mazeika, and Thomas Dietterich. Deep anomaly detection with outlier exposure. In International Conference on Learning Representations, 2019.

Yen-Chang Hsu, Yilin Shen, Hongxia Jin, and Zsolt Kira. Generalized ODIN: Detecting out-of distribution image without learning from out-of-distribution data. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020, pp. 10951-10960.

Rui Huang and Yixuan Li. Mos: Towards scaling out-of-distribution detection for large semantic space. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2021.

Ziqian Lin, Sreya Dutta, and Yixuan Li. Mood: Multi-level out-of-distribution detection. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2021.

Yann LeCun, Sumit Chopra, Raia Hadsell, Marc’Aurelio Ranzato, and Fu-Jie Huang. A tutorial on energy-based learning. In G. Bakir, T. Hofman, B. Schölkopf, A. Smola, and B. Taskar, editors, Predicting Structured Data. MIT Press, 2006.

Jinsong Zhang, Qiang Fu, Xu Chen, Lun Du, Zelin Li, Gang Wang, xiaoguang Liu, Shi Han, Dongmei Zhang, Out-of-Distribution Detection based on In-Distribution Data Patterns Memorization with Modern Hopfield Energy, ICLR 2023 Conference Program Chairs, 2023.