Towards Rigorous Interpretations: a Formalisation of Feature Attribution

Darius Afchar, Vincent Guigue, Romain Hennequin
Proceedings of the 38th International Conference on Machine Learning, PMLR 139:76-86, 2021.

Abstract

Feature attribution is often loosely presented as the process of selecting a subset of relevant features as a rationale of a prediction. Task-dependent by nature, precise definitions of "relevance" encountered in the literature are however not always consistent. This lack of clarity stems from the fact that we usually do not have access to any notion of ground-truth attribution and from a more general debate on what good interpretations are. In this paper we propose to formalise feature selection/attribution based on the concept of relaxed functional dependence. In particular, we extend our notions to the instance-wise setting and derive necessary properties for candidate selection solutions, while leaving room for task-dependence. By computing ground-truth attributions on synthetic datasets, we evaluate many state-of-the-art attribution methods and show that, even when optimised, some fail to verify the proposed properties and provide wrong solutions.

Cite this Paper


BibTeX
@InProceedings{pmlr-v139-afchar21a, title = {Towards Rigorous Interpretations: a Formalisation of Feature Attribution}, author = {Afchar, Darius and Guigue, Vincent and Hennequin, Romain}, booktitle = {Proceedings of the 38th International Conference on Machine Learning}, pages = {76--86}, year = {2021}, editor = {Meila, Marina and Zhang, Tong}, volume = {139}, series = {Proceedings of Machine Learning Research}, month = {18--24 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v139/afchar21a/afchar21a.pdf}, url = {https://proceedings.mlr.press/v139/afchar21a.html}, abstract = {Feature attribution is often loosely presented as the process of selecting a subset of relevant features as a rationale of a prediction. Task-dependent by nature, precise definitions of "relevance" encountered in the literature are however not always consistent. This lack of clarity stems from the fact that we usually do not have access to any notion of ground-truth attribution and from a more general debate on what good interpretations are. In this paper we propose to formalise feature selection/attribution based on the concept of relaxed functional dependence. In particular, we extend our notions to the instance-wise setting and derive necessary properties for candidate selection solutions, while leaving room for task-dependence. By computing ground-truth attributions on synthetic datasets, we evaluate many state-of-the-art attribution methods and show that, even when optimised, some fail to verify the proposed properties and provide wrong solutions.} }
Endnote
%0 Conference Paper %T Towards Rigorous Interpretations: a Formalisation of Feature Attribution %A Darius Afchar %A Vincent Guigue %A Romain Hennequin %B Proceedings of the 38th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2021 %E Marina Meila %E Tong Zhang %F pmlr-v139-afchar21a %I PMLR %P 76--86 %U https://proceedings.mlr.press/v139/afchar21a.html %V 139 %X Feature attribution is often loosely presented as the process of selecting a subset of relevant features as a rationale of a prediction. Task-dependent by nature, precise definitions of "relevance" encountered in the literature are however not always consistent. This lack of clarity stems from the fact that we usually do not have access to any notion of ground-truth attribution and from a more general debate on what good interpretations are. In this paper we propose to formalise feature selection/attribution based on the concept of relaxed functional dependence. In particular, we extend our notions to the instance-wise setting and derive necessary properties for candidate selection solutions, while leaving room for task-dependence. By computing ground-truth attributions on synthetic datasets, we evaluate many state-of-the-art attribution methods and show that, even when optimised, some fail to verify the proposed properties and provide wrong solutions.
APA
Afchar, D., Guigue, V. & Hennequin, R.. (2021). Towards Rigorous Interpretations: a Formalisation of Feature Attribution. Proceedings of the 38th International Conference on Machine Learning, in Proceedings of Machine Learning Research 139:76-86 Available from https://proceedings.mlr.press/v139/afchar21a.html.

Related Material