Marginal Contribution Feature Importance - an Axiomatic Approach for Explaining Data

Amnon Catav, Boyang Fu, Yazeed Zoabi, Ahuva Libi Weiss Meilik, Noam Shomron, Jason Ernst, Sriram Sankararaman, Ran Gilad-Bachrach
Proceedings of the 38th International Conference on Machine Learning, PMLR 139:1324-1335, 2021.

Abstract

In recent years, methods were proposed for assigning feature importance scores to measure the contribution of individual features. While in some cases the goal is to understand a specific model, in many cases the goal is to understand the contribution of certain properties (features) to a real-world phenomenon. Thus, a distinction has been made between feature importance scores that explain a model and scores that explain the data. When explaining the data, machine learning models are used as proxies in settings where conducting many real-world experiments is expensive or prohibited. While existing feature importance scores show great success in explaining models, we demonstrate their limitations when explaining the data, especially in the presence of correlations between features. Therefore, we develop a set of axioms to capture properties expected from a feature importance score when explaining data and prove that there exists only one score that satisfies all of them, the Marginal Contribution Feature Importance (MCI). We analyze the theoretical properties of this score function and demonstrate its merits empirically.

Cite this Paper


BibTeX
@InProceedings{pmlr-v139-catav21a, title = {Marginal Contribution Feature Importance - an Axiomatic Approach for Explaining Data}, author = {Catav, Amnon and Fu, Boyang and Zoabi, Yazeed and Meilik, Ahuva Libi Weiss and Shomron, Noam and Ernst, Jason and Sankararaman, Sriram and Gilad-Bachrach, Ran}, booktitle = {Proceedings of the 38th International Conference on Machine Learning}, pages = {1324--1335}, 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/catav21a/catav21a.pdf}, url = {https://proceedings.mlr.press/v139/catav21a.html}, abstract = {In recent years, methods were proposed for assigning feature importance scores to measure the contribution of individual features. While in some cases the goal is to understand a specific model, in many cases the goal is to understand the contribution of certain properties (features) to a real-world phenomenon. Thus, a distinction has been made between feature importance scores that explain a model and scores that explain the data. When explaining the data, machine learning models are used as proxies in settings where conducting many real-world experiments is expensive or prohibited. While existing feature importance scores show great success in explaining models, we demonstrate their limitations when explaining the data, especially in the presence of correlations between features. Therefore, we develop a set of axioms to capture properties expected from a feature importance score when explaining data and prove that there exists only one score that satisfies all of them, the Marginal Contribution Feature Importance (MCI). We analyze the theoretical properties of this score function and demonstrate its merits empirically.} }
Endnote
%0 Conference Paper %T Marginal Contribution Feature Importance - an Axiomatic Approach for Explaining Data %A Amnon Catav %A Boyang Fu %A Yazeed Zoabi %A Ahuva Libi Weiss Meilik %A Noam Shomron %A Jason Ernst %A Sriram Sankararaman %A Ran Gilad-Bachrach %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-catav21a %I PMLR %P 1324--1335 %U https://proceedings.mlr.press/v139/catav21a.html %V 139 %X In recent years, methods were proposed for assigning feature importance scores to measure the contribution of individual features. While in some cases the goal is to understand a specific model, in many cases the goal is to understand the contribution of certain properties (features) to a real-world phenomenon. Thus, a distinction has been made between feature importance scores that explain a model and scores that explain the data. When explaining the data, machine learning models are used as proxies in settings where conducting many real-world experiments is expensive or prohibited. While existing feature importance scores show great success in explaining models, we demonstrate their limitations when explaining the data, especially in the presence of correlations between features. Therefore, we develop a set of axioms to capture properties expected from a feature importance score when explaining data and prove that there exists only one score that satisfies all of them, the Marginal Contribution Feature Importance (MCI). We analyze the theoretical properties of this score function and demonstrate its merits empirically.
APA
Catav, A., Fu, B., Zoabi, Y., Meilik, A.L.W., Shomron, N., Ernst, J., Sankararaman, S. & Gilad-Bachrach, R.. (2021). Marginal Contribution Feature Importance - an Axiomatic Approach for Explaining Data. Proceedings of the 38th International Conference on Machine Learning, in Proceedings of Machine Learning Research 139:1324-1335 Available from https://proceedings.mlr.press/v139/catav21a.html.

Related Material