nnitp: Neural net interpretability with Bayesian interpolants¶
Nnitp is a tool for interpretability of neural network inferences using Bayesian interpolants. The basic technique is described in this paper. It currently supports convolutional neural networks for image classification, using the Keras framework.