Assurance for Learning-Enabled Systems (ALES)
Abstract
The Assurance for Learning-Enabled Systems project develops an integrated well-founded approach to the co-design of cyberphysical systems with learning-enabled components and their assurance arguments for both perception and control. The approach was based on a theory of mind-predictive processing (PP) complemented with the dual process theories of reasoning. The research developed a two-layered runtime monitor using generative modeling in the first layer and graph Markov neural networks based neurosymbolic contextual modeling in the second layer.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 01, 2022
- Accession Number
- AD1170176
Entities
People
- Susmit Jha
Organizations
- SRI International