EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

An Empirical Comparison of Interpretable Models to Post-Hoc Explanations

Parisa Mahya, Johannes Fürnkranz

AI · 19.05.2023 · 4 (2) · 426-436

DOI: 10.3390/ai4020023 ↗

Kurz gefasst

Fachartikel in der Zeitschrift AI, online erschienen am 19.05.2023. Verfasst von Parisa Mahya und 1 weiteren Autor:innen. Die Arbeit wurde bisher 15-mal zitiert (Crossref) und nennt 22 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Recently, some effort went into explaining intransparent and black-box models, such as deep neural networks or random forests. So-called model-agnostic methods typically approximate the prediction of the intransparent black-box model with an interpretable model, without considering any specifics of the black-box model itself. It is a valid question whether direct learning of interpretable white-box models should not be preferred over post-hoc approximations of intransparent and black-box models. In this paper, we report the results of an empirical study, which compares post-hoc explanations and interpretable models on several datasets for rule-based and feature-based interpretable models. The results seem to underline that often directly learned interpretable models approximate the black-box models at least as well as their post-hoc surrogates, even though the former do not have direct access to the black-box model.

Zitieren

APA

Mahya, P., & Fürnkranz, J. (2023). An Empirical Comparison of Interpretable Models to Post-Hoc Explanations. AI, 4(2), 426-436. https://doi.org/10.3390/ai4020023

Vancouver

Mahya P, Fürnkranz J. An Empirical Comparison of Interpretable Models to Post-Hoc Explanations. AI. 2023;4(2):426-436. doi:10.3390/ai4020023

BibTeX
@article{Mahya2023,
  author = {Mahya, Parisa and Fürnkranz, Johannes},
  title = {An Empirical Comparison of Interpretable Models to Post-Hoc Explanations},
  journal = {AI},
  year = {2023},
  volume = {4},
  number = {2},
  pages = {426--436},
  doi = {10.3390/ai4020023}
}
Kostenlos starten →