Safe Predictors for Enforcing Input-Output Specifications
This addresses the critical need for reliable and verifiable AI systems in safety-critical domains like aviation, though it appears incremental as it builds on existing constraint-based methods.
The paper tackles the problem of designing neural networks that are guaranteed to satisfy input-output specifications throughout training and deployment, by developing a method that combines constrained predictors via convex combinations. The approach is demonstrated on synthetic datasets and an aircraft collision avoidance problem.
We present an approach for designing correct-by-construction neural networks (and other machine learning models) that are guaranteed to be consistent with a collection of input-output specifications before, during, and after algorithm training. Our method involves designing a constrained predictor for each set of compatible constraints, and combining them safely via a convex combination of their predictions. We demonstrate our approach on synthetic datasets and an aircraft collision avoidance problem.