parametric t-SNE

New method linearizes neural networks for enhanced interpretability

A new research paper proposes an instance-wise linearization approach to improve the interpretability of neural networks. The method reformulates the forward computation of a neural network into a linear matrix multipli…