Probably Approximately Correct Explanations of Machine Learning Models via Syntax-Guided Synthesis
This addresses the need for interpretable AI explanations, though it appears incremental as it builds on existing PAC and SyGuS methods.
The paper tackles the problem of explaining complex machine learning models by proposing a framework that combines probably approximately correct learning and syntax-guided synthesis, proving it generates explanations with high probability of few errors and showing empirically it produces small, human-interpretable results.
We propose a novel approach to understanding the decision making of complex machine learning models (e.g., deep neural networks) using a combination of probably approximately correct learning (PAC) and a logic inference methodology called syntax-guided synthesis (SyGuS). We prove that our framework produces explanations that with a high probability make only few errors and show empirically that it is effective in generating small, human-interpretable explanations.