Summary of Vehicle: Bridging the Embedding Gap in the Verification Of Neuro-symbolic Programs, by Matthew L. Daggitt et al.
Vehicle: Bridging the Embedding Gap in the Verification of Neuro-Symbolic Programs
by Matthew L. Daggitt, Wen Kokke, Robert Atkey, Natalia Slusarz, Luca Arnaboldi, Ekaterina Komendantskaya
First submitted to arxiv on: 12 Jan 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- Secondary: None
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 addresses the lack of methodology for verifying neuro-symbolic programs that rely on machine learning components. The authors identify the “embedding gap” as a key issue, where semantically meaningful properties in the problem space need to be linked to equivalent properties in the embedding space. To bridge this gap, they introduce Vehicle, a tool for end-to-end verification of neural-symbolic programs in a modular fashion. Vehicle provides a language for specifying problem-space properties and their relationship to the embedding space, as well as a compiler that automates interpretation of these properties in various machine-learning training environments, verifiers, and theorem provers. The authors demonstrate Vehicle’s utility by formally verifying the safety of a simple autonomous car with a neural network controller. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about making sure computer programs are correct when they use both old-style code and new-style machine learning tools. They found that this mix of old and new is hard to check, so they created a tool called Vehicle to help with this problem. Vehicle lets you describe the rules for how a program should behave and then translates those rules into the language of different machines that can do the checking. The authors show that Vehicle works by using it to prove that a simple self-driving car’s neural network controller is safe. |
Keywords
* Artificial intelligence * Embedding * Embedding space * Machine learning * Neural network
