Summary of Taming False Positives in Out-of-distribution Detection with Human Feedback, by Harit Vishwakarma et al.
Taming False Positives in Out-of-Distribution Detection with Human Feedback
by Harit Vishwakarma, Heguang Lin, Ramya Korlakai Vinayak
First submitted to arxiv on: 25 Apr 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
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 Recent works have focused on designing scoring functions to quantify out-of-distribution (OOD) uncertainty in machine learning models. However, setting thresholds for these scoring functions is challenging due to the lack of OOD samples upfront. Typically, thresholds are set to achieve a desired true positive rate (TPR), but this can lead to high false positive rates (FPR). To address this, we propose a mathematically grounded OOD detection framework that leverages expert feedback to update the threshold on the fly. Our framework ensures it meets the FPR constraint at all times while minimizing human feedback. We also show that our method works with any scoring function for OOD uncertainty quantification. Empirically, our system maintains FPR at most 5% while maximizing TPR on synthetic and benchmark datasets. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Imagine you’re trying to teach a machine learning model to detect when it’s being given information that’s not normal or typical. This is important because in the real world, machines don’t always get perfect data. Some recent papers have focused on creating special scores to measure how unusual something is. But setting limits for these scores is tricky because we often don’t know ahead of time what’s considered “out-of-the-ordinary.” Our new approach uses expert feedback to adjust the score limits as needed, making sure they stay within a safe range. We also show that our method can work with different scoring systems. In tests, our system did well at detecting unusual data while keeping false alarms low. |
Keywords
» Artificial intelligence » Machine learning
