Efficient SAGE Estimation via Causal Structure Learning

Christoph Luther, Gunnar König, Moritz Grosse-Wentrup
Proceedings of The 26th International Conference on Artificial Intelligence and Statistics, PMLR 206:11650-11670, 2023.

Abstract

The Shapley Additive Global Importance (SAGE) value is a theoretically appealing interpretability method that fairly attributes global importance to a model’s surplus performance contributions over an exponential number of feature sets. This is computationally expensive, particularly because estimating the surplus contributions requires sampling from conditional distributions. Thus, SAGE approximation algorithms only take a fraction of the feature sets into account. We propose d-SAGE, a method that accelerates SAGE approximation. d-SAGE is motivated by the observation that conditional independencies (CIs) between a feature and the model target imply zero surplus contributions, such that their computation can be skipped. To identify CIs, we leverage causal structure learning (CSL) to infer a graph that encodes (conditional) independencies in the data as d-separations. This is computationally more efficient because the expense of the one-time graph inference and the d-separation queries is negligible compared to the expense of surplus contribution evaluations. Empirically we demonstrate that d-SAGE enables the efficient and accurate estimation of SAGE values.

Cite this Paper

BibTeX
@InProceedings{pmlr-v206-luther23a, title = {Efficient SAGE Estimation via Causal Structure Learning}, author = {Luther, Christoph and K\"onig, Gunnar and Grosse-Wentrup, Moritz}, booktitle = {Proceedings of The 26th International Conference on Artificial Intelligence and Statistics}, pages = {11650--11670}, year = {2023}, editor = {Ruiz, Francisco and Dy, Jennifer and van de Meent, Jan-Willem}, volume = {206}, series = {Proceedings of Machine Learning Research}, month = {25--27 Apr}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v206/luther23a/luther23a.pdf}, url = {https://proceedings.mlr.press/v206/luther23a.html}, abstract = {The Shapley Additive Global Importance (SAGE) value is a theoretically appealing interpretability method that fairly attributes global importance to a model’s surplus performance contributions over an exponential number of feature sets. This is computationally expensive, particularly because estimating the surplus contributions requires sampling from conditional distributions. Thus, SAGE approximation algorithms only take a fraction of the feature sets into account. We propose d-SAGE, a method that accelerates SAGE approximation. d-SAGE is motivated by the observation that conditional independencies (CIs) between a feature and the model target imply zero surplus contributions, such that their computation can be skipped. To identify CIs, we leverage causal structure learning (CSL) to infer a graph that encodes (conditional) independencies in the data as d-separations. This is computationally more efficient because the expense of the one-time graph inference and the d-separation queries is negligible compared to the expense of surplus contribution evaluations. Empirically we demonstrate that d-SAGE enables the efficient and accurate estimation of SAGE values.} }
Endnote
%0 Conference Paper %T Efficient SAGE Estimation via Causal Structure Learning %A Christoph Luther %A Gunnar König %A Moritz Grosse-Wentrup %B Proceedings of The 26th International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2023 %E Francisco Ruiz %E Jennifer Dy %E Jan-Willem van de Meent %F pmlr-v206-luther23a %I PMLR %P 11650--11670 %U https://proceedings.mlr.press/v206/luther23a.html %V 206 %X The Shapley Additive Global Importance (SAGE) value is a theoretically appealing interpretability method that fairly attributes global importance to a model’s surplus performance contributions over an exponential number of feature sets. This is computationally expensive, particularly because estimating the surplus contributions requires sampling from conditional distributions. Thus, SAGE approximation algorithms only take a fraction of the feature sets into account. We propose d-SAGE, a method that accelerates SAGE approximation. d-SAGE is motivated by the observation that conditional independencies (CIs) between a feature and the model target imply zero surplus contributions, such that their computation can be skipped. To identify CIs, we leverage causal structure learning (CSL) to infer a graph that encodes (conditional) independencies in the data as d-separations. This is computationally more efficient because the expense of the one-time graph inference and the d-separation queries is negligible compared to the expense of surplus contribution evaluations. Empirically we demonstrate that d-SAGE enables the efficient and accurate estimation of SAGE values.
APA
Luther, C., König, G. & Grosse-Wentrup, M.. (2023). Efficient SAGE Estimation via Causal Structure Learning. Proceedings of The 26th International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 206:11650-11670 Available from https://proceedings.mlr.press/v206/luther23a.html.

Related Material