We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.LG

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Machine Learning

Title: ML Interpretability: Simple Isn't Easy

Authors: Tim Räz
Abstract: The interpretability of ML models is important, but it is not clear what it amounts to. So far, most philosophers have discussed the lack of interpretability of black-box models such as neural networks, and methods such as explainable AI that aim to make these models more transparent. The goal of this paper is to clarify the nature of interpretability by focussing on the other end of the 'interpretability spectrum'. The reasons why some models, linear models and decision trees, are highly interpretable will be examined, and also how more general models, MARS and GAM, retain some degree of interpretability. I find that while there is heterogeneity in how we gain interpretability, what interpretability is in particular cases can be explicated in a clear manner.
Comments: 22 pages, 4 figures
Subjects: Machine Learning (cs.LG); Computers and Society (cs.CY)
Cite as: arXiv:2211.13617 [cs.LG]
  (or arXiv:2211.13617v1 [cs.LG] for this version)

Submission history

From: Tim Räz [view email]
[v1] Thu, 24 Nov 2022 14:07:02 GMT (842kb,D)

Link back to: arXiv, form interface, contact.