Summary of Fun-ad: Fully Unsupervised Learning For Anomaly Detection with Noisy Training Data, by Jiin Im et al.
FUN-AD: Fully Unsupervised Learning for Anomaly Detection with Noisy Training Data
by Jiin Im, Yongho Son, Je Hyeong Hong
First submitted to arxiv on: 25 Nov 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 A novel, fully unsupervised anomaly detection approach for handling noisy training data in industrial environments is proposed. The method leverages pairwise feature distances and mutual closeness to identify normal samples and anomalies. A pseudo-labeling strategy uses an iteratively reconstructed memory bank (IRMB) to distinguish confident normal samples from anomalies. A new loss function promotes class-homogeneity, reducing task ill-posedness. Experimental results on two public benchmarks and semantic examples validate FUN-AD’s effectiveness across various scenarios. The code is available at https://github.com/HY-Vision-Lab/FUNAD. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary A group of scientists developed a new way to find unusual things in a dataset without knowing what most of the data looks like. This is useful when there are mistakes in the training data or not enough information for certain products. The method uses two ideas: first, it’s easier to tell normal samples from anomalies if you know how close they are; and second, pairs of features that are very similar are likely to be from the same group. They created a strategy using this idea and called it pseudo-labeling. It helps identify confident normal samples and anomalies. The scientists also created a new way to measure how well their method works by looking at how well similar things are grouped together. They tested their method on real-world data and showed that it can work well even when there are different ratios of unusual to normal things. |
Keywords
» Artificial intelligence » Anomaly detection » Loss function » Unsupervised
