Summary of Neural Control and Certificate Repair Via Runtime Monitoring, by Emily Yu et al.
Neural Control and Certificate Repair via Runtime Monitoring
by Emily Yu, Đorđe Žikelić, Thomas A. Henzinger
First submitted to arxiv on: 17 Dec 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI)
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 framework uses runtime monitoring to detect system behaviors that violate a safety property under an initially trained neural network policy and certificate. This is done in the black-box setting where the system dynamics are unknown, unlike previous work which focused on white-box verification. The approach involves extracting new training data from violating behaviors, re-training the neural network policy and certificate function, and ultimately repairing them to improve their safety rate. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper explores ways to ensure that a control policy learned by a computer stays safe as it interacts with an unknown environment. Currently, most work on this topic assumes that we can see what’s happening inside the computer (white-box), but in reality, many systems are black boxes where we don’t have direct access. The researchers developed a new method to monitor and improve the safety of control policies learned by neural networks in these black box situations. They tested their approach on two real-world problems and showed that it can effectively repair and improve the safety of control policies. |
Keywords
» Artificial intelligence » Neural network
