Neuron Activation-based Computation of Logical Explanations for Deep Neural Networks
This paper addresses formal explainability of classifying neural networks by introducing a flexible symbolic framework for an efficient, guided computation of explanations of the NN behavior, parametrized by the activations of internal neurons, and using logical engines such as SMT solvers.
Tomáš Kolárik, Faezeh Labbaf, Fabrizio Leopardi et al.
· 0 citations