Summary of Improving Out-of-distribution Detection by Combining Existing Post-hoc Methods, By Paul Novello et al.
Improving Out-of-Distribution Detection by Combining Existing Post-hoc Methods
by Paul Novello, Yannick Prudent, Joseba Dalmau, Corentin Friedrich, Yann Pequignot
First submitted to arxiv on: 9 Jul 2024
Categories
- Main: Machine Learning (stat.ML)
- Secondary: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
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 proposed paper focuses on developing a framework for effectively combining existing Out-of-Distribution (OOD) detection methods to enhance OOD detection. The authors propose and compare four different strategies for integrating multiple detection scores into a unified OOD detector, utilizing techniques such as majority vote, cumulative distribution function modeling, and multivariate quantiles based on optimal transport. To evaluate these multi-dimensional OOD detectors, the authors extend common OOD evaluation metrics like AUROC and FPR at fixed TPR rates to these combined methods. The paper also provides guidelines for choosing which OOD detectors to combine in more realistic settings without known OOD data, drawing from principles of Outlier Exposure. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper explores ways to combine different Out-of-Distribution (OOD) detection methods to improve robustness in safety-critical applications. Instead of developing new methods, the authors focus on integrating existing ones using techniques like majority vote and cumulative distribution function modeling. The goal is to create a unified OOD detector that can perform well across various datasets. |
