Prediction via Shapley Value Regression

Amr Alkhatib, Roman Bresson, Henrik Boström, Michalis Vazirgiannis
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:1056-1101, 2025.

Abstract

Shapley values have several desirable, theoretically well-supported, properties for explaining black-box model predictions. Traditionally, Shapley values are computed post-hoc, leading to additional computational cost at inference time. To overcome this, a novel method, called ViaSHAP, is proposed, that learns a function to compute Shapley values, from which the predictions can be derived directly by summation. Two approaches to implement the proposed method are explored; one based on the universal approximation theorem and the other on the Kolmogorov-Arnold representation theorem. Results from a large-scale empirical investigation are presented, showing that ViaSHAP using Kolmogorov-Arnold Networks performs on par with state-of-the-art algorithms for tabular data. It is also shown that the explanations of ViaSHAP are significantly more accurate than the popular approximator FastSHAP on both tabular data and images.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-alkhatib25a, title = {Prediction via Shapley Value Regression}, author = {Alkhatib, Amr and Bresson, Roman and Bostr\"{o}m, Henrik and Vazirgiannis, Michalis}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {1056--1101}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/alkhatib25a/alkhatib25a.pdf}, url = {https://proceedings.mlr.press/v267/alkhatib25a.html}, abstract = {Shapley values have several desirable, theoretically well-supported, properties for explaining black-box model predictions. Traditionally, Shapley values are computed post-hoc, leading to additional computational cost at inference time. To overcome this, a novel method, called ViaSHAP, is proposed, that learns a function to compute Shapley values, from which the predictions can be derived directly by summation. Two approaches to implement the proposed method are explored; one based on the universal approximation theorem and the other on the Kolmogorov-Arnold representation theorem. Results from a large-scale empirical investigation are presented, showing that ViaSHAP using Kolmogorov-Arnold Networks performs on par with state-of-the-art algorithms for tabular data. It is also shown that the explanations of ViaSHAP are significantly more accurate than the popular approximator FastSHAP on both tabular data and images.} }
Endnote
%0 Conference Paper %T Prediction via Shapley Value Regression %A Amr Alkhatib %A Roman Bresson %A Henrik Boström %A Michalis Vazirgiannis %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-alkhatib25a %I PMLR %P 1056--1101 %U https://proceedings.mlr.press/v267/alkhatib25a.html %V 267 %X Shapley values have several desirable, theoretically well-supported, properties for explaining black-box model predictions. Traditionally, Shapley values are computed post-hoc, leading to additional computational cost at inference time. To overcome this, a novel method, called ViaSHAP, is proposed, that learns a function to compute Shapley values, from which the predictions can be derived directly by summation. Two approaches to implement the proposed method are explored; one based on the universal approximation theorem and the other on the Kolmogorov-Arnold representation theorem. Results from a large-scale empirical investigation are presented, showing that ViaSHAP using Kolmogorov-Arnold Networks performs on par with state-of-the-art algorithms for tabular data. It is also shown that the explanations of ViaSHAP are significantly more accurate than the popular approximator FastSHAP on both tabular data and images.
APA
Alkhatib, A., Bresson, R., Boström, H. & Vazirgiannis, M.. (2025). Prediction via Shapley Value Regression. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:1056-1101 Available from https://proceedings.mlr.press/v267/alkhatib25a.html.

Related Material