Summary of Can We Defend Against the Unknown? An Empirical Study About Threshold Selection For Neural Network Monitoring, by Khoi Tran Dang et al.
Can we Defend Against the Unknown? An Empirical Study About Threshold Selection for Neural Network Monitoring
by Khoi Tran Dang, Kevin Delmas, Jérémie Guiochet, Joris Guérin
First submitted to arxiv on: 14 May 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Image and Video Processing (eess.IV)
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 This paper tackles the crucial issue of runtime monitoring in neural networks used in critical systems. Existing methods focus on rejecting unsafe predictions, but overlook the importance of setting an effective threshold to transform scores into actionable decisions. The authors argue that current approaches assume a strong relationship between training and runtime data distributions, which is unrealistic given the unpredictability of unforeseen threats. They conduct experiments on various image datasets to investigate the effectiveness of monitors in handling unexpected threats and whether integrating generic threats into the threshold optimization process can enhance robustness. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This research paper looks at how we can make sure that artificial intelligence systems don’t make mistakes when they’re making decisions. Right now, people are using neural networks to make predictions, but it’s hard to know if those predictions are correct or not. The authors think that the way we do this is important because it affects how we handle unexpected problems. They tested different methods on image datasets to see what works best and how we can make our decisions more reliable. |
Keywords
» Artificial intelligence » Optimization
