Interpretable Neural Networks with Frank-Wolfe: Sparse Relevance Maps and Relevance Orderings

Jan Macdonald, Mathieu E. Besançon, Sebastian Pokutta
Proceedings of the 39th International Conference on Machine Learning, PMLR 162:14699-14716, 2022.

Abstract

We study the effects of constrained optimization formulations and Frank-Wolfe algorithms for obtaining interpretable neural network predictions. Reformulating the Rate-Distortion Explanations (RDE) method for relevance attribution as a constrained optimization problem provides precise control over the sparsity of relevance maps. This enables a novel multi-rate as well as a relevance-ordering variant of RDE that both empirically outperform standard RDE and other baseline methods in a well-established comparison test. We showcase several deterministic and stochastic variants of the Frank-Wolfe algorithm and their effectiveness for RDE.

Cite this Paper

BibTeX
@InProceedings{pmlr-v162-macdonald22a, title = {Interpretable Neural Networks with Frank-Wolfe: Sparse Relevance Maps and Relevance Orderings}, author = {Macdonald, Jan and Besan{\c{c}}on, Mathieu E. and Pokutta, Sebastian}, booktitle = {Proceedings of the 39th International Conference on Machine Learning}, pages = {14699--14716}, year = {2022}, editor = {Chaudhuri, Kamalika and Jegelka, Stefanie and Song, Le and Szepesvari, Csaba and Niu, Gang and Sabato, Sivan}, volume = {162}, series = {Proceedings of Machine Learning Research}, month = {17--23 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v162/macdonald22a/macdonald22a.pdf}, url = {https://proceedings.mlr.press/v162/macdonald22a.html}, abstract = {We study the effects of constrained optimization formulations and Frank-Wolfe algorithms for obtaining interpretable neural network predictions. Reformulating the Rate-Distortion Explanations (RDE) method for relevance attribution as a constrained optimization problem provides precise control over the sparsity of relevance maps. This enables a novel multi-rate as well as a relevance-ordering variant of RDE that both empirically outperform standard RDE and other baseline methods in a well-established comparison test. We showcase several deterministic and stochastic variants of the Frank-Wolfe algorithm and their effectiveness for RDE.} }
Endnote
%0 Conference Paper %T Interpretable Neural Networks with Frank-Wolfe: Sparse Relevance Maps and Relevance Orderings %A Jan Macdonald %A Mathieu E. Besançon %A Sebastian Pokutta %B Proceedings of the 39th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2022 %E Kamalika Chaudhuri %E Stefanie Jegelka %E Le Song %E Csaba Szepesvari %E Gang Niu %E Sivan Sabato %F pmlr-v162-macdonald22a %I PMLR %P 14699--14716 %U https://proceedings.mlr.press/v162/macdonald22a.html %V 162 %X We study the effects of constrained optimization formulations and Frank-Wolfe algorithms for obtaining interpretable neural network predictions. Reformulating the Rate-Distortion Explanations (RDE) method for relevance attribution as a constrained optimization problem provides precise control over the sparsity of relevance maps. This enables a novel multi-rate as well as a relevance-ordering variant of RDE that both empirically outperform standard RDE and other baseline methods in a well-established comparison test. We showcase several deterministic and stochastic variants of the Frank-Wolfe algorithm and their effectiveness for RDE.
APA
Macdonald, J., Besançon, M.E. & Pokutta, S.. (2022). Interpretable Neural Networks with Frank-Wolfe: Sparse Relevance Maps and Relevance Orderings. Proceedings of the 39th International Conference on Machine Learning, in Proceedings of Machine Learning Research 162:14699-14716 Available from https://proceedings.mlr.press/v162/macdonald22a.html.

Related Material