Daily Arxiv

This is a page that curates AI-related papers published worldwide.
All content here is summarized using Google Gemini and operated on a non-profit basis.
Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.

Unsupervised Anomaly Detection through Mass Repulsing Optimal Transport

Created by
  • Haebom

Author

Eduardo Fernandes Montesuma, Adel El Habazi, Fred Ngole Mboula

Outline

This paper proposes Mass Repulsing Optimal Transport (MROT), a new method for outlier detection, by utilizing the existing Optimal Transport (OT) theory. MROT is a method that moves the mass of data with minimal effort, while forcing samples in low-density areas suspected of being outliers to move their mass farther away, incurring high transport costs. We utilize this transport cost as an outlier score to detect outliers, and demonstrate through experiments that it outperforms existing outlier detection methods.

Takeaways, Limitations

•
Takeaways:
◦
We present a novel outlier detection algorithm that demonstrates improved performance over existing outlier detection methods.
◦
A case study of effective application of optimal transportation theory to the problem of outlier detection.
◦
Performance validation on various experimental datasets (existing benchmarks and fault detection problems).
•
Limitations:
◦
Lack of analysis on the computational complexity and scalability of the proposed MROT algorithm.
◦
Further research is needed on generalization performance for datasets of different dimensions and sizes.
◦
Possible performance degradation for certain types of outliers and lack of analysis for them.
👍
Made with Slashpage