Bayesian Interpolants as Explanations for Neural Inferences
arXiv:2004.04198v12 citations
AI Analysis
This addresses the interpretability problem for users of neural networks, though it appears incremental as it adapts an existing concept to a new domain.
The paper tackles the problem of explaining neural network inferences by adapting Craig interpolants from logical to statistical inference, resulting in explanations that are concise, understandable, and precise.
The notion of Craig interpolant, used as a form of explanation in automated reasoning, is adapted from logical inference to statistical inference and used to explain inferences made by neural networks. The method produces explanations that are at the same time concise, understandable and precise.