It is widely recognized that the predictions of deep neural networks are difficult to parse relative to simpler approaches. However, the development of methods to investigate the mode of operation of such models has advanced rapidly in the past few years. Recent work introduced an intuitive framework which utilizes generative models to improve on the meaningfulness of such explanations. In this work, we display the flexibility of this method to interpret diverse and challenging modalities: music and physical simulations of urban environments.
@article{arxiv.2007.00758,
title = {In-Distribution Interpretability for Challenging Modalities},
author = {Cosmas Heiß and Ron Levie and Cinjon Resnick and Gitta Kutyniok and Joan Bruna},
journal= {arXiv preprint arXiv:2007.00758},
year = {2020}
}