Explanations for Monotonic Classifiers.

Joao Marques-Silva, Thomas Gerspacher, Martin C Cooper, Alexey Ignatiev, Nina Narodytska
Proceedings of the 38th International Conference on Machine Learning, PMLR 139:7469-7479, 2021.

Abstract

In many classification tasks there is a requirement of monotonicity. Concretely, if all else remains constant, increasing (resp. decreasing) the value of one or more features must not decrease (resp. increase) the value of the prediction. Despite comprehensive efforts on learning monotonic classifiers, dedicated approaches for explaining monotonic classifiers are scarce and classifier-specific. This paper describes novel algorithms for the computation of one formal explanation of a (black-box) monotonic classifier. These novel algorithms are polynomial (indeed linear) in the run time complexity of the classifier. Furthermore, the paper presents a practically efficient model-agnostic algorithm for enumerating formal explanations.

Cite this Paper

BibTeX
@InProceedings{pmlr-v139-marques-silva21a, title = {Explanations for Monotonic Classifiers.}, author = {Marques-Silva, Joao and Gerspacher, Thomas and Cooper, Martin C and Ignatiev, Alexey and Narodytska, Nina}, booktitle = {Proceedings of the 38th International Conference on Machine Learning}, pages = {7469--7479}, 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/marques-silva21a/marques-silva21a.pdf}, url = {https://proceedings.mlr.press/v139/marques-silva21a.html}, abstract = {In many classification tasks there is a requirement of monotonicity. Concretely, if all else remains constant, increasing (resp. decreasing) the value of one or more features must not decrease (resp. increase) the value of the prediction. Despite comprehensive efforts on learning monotonic classifiers, dedicated approaches for explaining monotonic classifiers are scarce and classifier-specific. This paper describes novel algorithms for the computation of one formal explanation of a (black-box) monotonic classifier. These novel algorithms are polynomial (indeed linear) in the run time complexity of the classifier. Furthermore, the paper presents a practically efficient model-agnostic algorithm for enumerating formal explanations.} }
Endnote
%0 Conference Paper %T Explanations for Monotonic Classifiers. %A Joao Marques-Silva %A Thomas Gerspacher %A Martin C Cooper %A Alexey Ignatiev %A Nina Narodytska %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-marques-silva21a %I PMLR %P 7469--7479 %U https://proceedings.mlr.press/v139/marques-silva21a.html %V 139 %X In many classification tasks there is a requirement of monotonicity. Concretely, if all else remains constant, increasing (resp. decreasing) the value of one or more features must not decrease (resp. increase) the value of the prediction. Despite comprehensive efforts on learning monotonic classifiers, dedicated approaches for explaining monotonic classifiers are scarce and classifier-specific. This paper describes novel algorithms for the computation of one formal explanation of a (black-box) monotonic classifier. These novel algorithms are polynomial (indeed linear) in the run time complexity of the classifier. Furthermore, the paper presents a practically efficient model-agnostic algorithm for enumerating formal explanations.
APA
Marques-Silva, J., Gerspacher, T., Cooper, M.C., Ignatiev, A. & Narodytska, N.. (2021). Explanations for Monotonic Classifiers.. Proceedings of the 38th International Conference on Machine Learning, in Proceedings of Machine Learning Research 139:7469-7479 Available from https://proceedings.mlr.press/v139/marques-silva21a.html.

Related Material