Summary of Hierarchical Gaussian Mixture Normalizing Flow Modeling For Unified Anomaly Detection, by Xincheng Yao and Ruoqi Li and Zefeng Qian and Lu Wang and Chongyang Zhang
Hierarchical Gaussian Mixture Normalizing Flow Modeling for Unified Anomaly Detection
by Xincheng Yao, Ruoqi Li, Zefeng Qian, Lu Wang, Chongyang Zhang
First submitted to arxiv on: 20 Mar 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Computer Vision and Pattern Recognition (cs.CV)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The paper proposes a novel approach to unified anomaly detection (AD) using hierarchical Gaussian mixture normalizing flow modeling, dubbed HGAD. The goal is to train a single model that can detect anomalies in multiple classes. Current NF-based AD methods may suffer from a “homogeneous mapping” issue, where they generate similar latent representations for both normal and abnormal features, leading to high missing rates of anomalies. HGAD addresses this by introducing inter-class Gaussian mixture modeling and intra-class mixed class centers learning. This approach can better represent complex multi-class distributions in the latent space, avoiding the “homogeneous mapping” issue. The method also incorporates a mutual information maximization loss to structure the latent feature space. Experimental results on four real-world AD benchmarks show significant improvement over previous NF-based AD methods and state-of-the-art unified AD methods. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary HGAD is a new way to detect anomalies in many different classes at once. This can be hard because existing methods might not be able to tell apart normal and abnormal features, which means they miss some of the anomalies. HGAD solves this problem by using a special kind of model that can handle many different classes and learn what makes each class unique. The method also helps the model avoid making all classes look the same, which is important for finding most of the anomalies. In tests on four real-world datasets, HGAD outperformed other methods at detecting anomalies. |
Keywords
* Artificial intelligence * Anomaly detection * Latent space
