A Functional Information Perspective on Model Interpretation
Published in ICML, 2022
Recommended citation: Itai Gat, Nitay Calderon, Roi Reichart and Tamir Hazan https://proceedings.mlr.press/v162/gat22a.html
Abstract
Contemporary predictive models are hard to interpret as their deep nets exploit numerous complex relations between input elements. This work suggests a theoretical framework for model interpretability by measuring the contribution of relevant features to the functional entropy of the network with respect to the input. We rely on the log-Sobolev inequality that bounds the functional entropy by the functional Fisher information with respect to the covariance of the data. This provides a principled way to measure the amount of information contribution of a subset of features to the decision function. Through extensive experiments, we show that our method surpasses existing interpretability sampling-based methods on various data signals such as image, text, and audio.bibtex
@inproceedings{gat2022functional, author = {Itai Gat and Nitay Calderon and Roi Reichart and Tamir Hazan}, editor = {Kamalika Chaudhuri and Stefanie Jegelka and Le Song and Csaba Szepesv{\'{a}}ri and Gang Niu and Sivan Sabato}, title = {A Functional Information Perspective on Model Interpretation}, booktitle = {International Conference on Machine Learning, {ICML} 2022, 17-23 July 2022, Baltimore, Maryland, {USA}}, series = {Proceedings of Machine Learning Research}, volume = {162}, pages = {7266--7278}, publisher = , year = {2022}, url = {https://proceedings.mlr.press/v162/gat22a.html}, timestamp = {Tue, 12 Jul 2022 17:36:52 +0200}, biburl = {https://dblp.org/rec/conf/icml/GatCRH22.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} }