Learning Qualitatively Diverse and Interpretable Rules for Classification

Andrew Slavin Ross, Weiwei Pan, Finale Doshi-Velez
6/22/2018
cs.LGstat.ML

Official code: dtak/local-independence-public (5 stars, MIT, last commit 2019-11-27).

Abstract

There has been growing interest in developing accurate models that can also be explained to humans. Unfortunately, if there exist multiple distinct but accurate models for some dataset, current machine learning methods are unlikely to find them: standard techniques will likely recover a complex model that combines them. In this work, we introduce a way to identify a maximal set of distinct but accurate models for a dataset. We demonstrate empirically that, in situations where the data supports multiple accurate classifiers, we tend to recover simpler, more interpretable classifiers rather than more complex ones.

View on arXivView PDF

Code Implementations(1)

dtak/local-independence-publicOfficial100%
MIT

Code/figures in Learning Qualitatively Diverse and Interpretable Rules for Classification

50Python, Jupyter NotebookJun 30, 20186 years agoMIT
interpretabilitymultiple-solutions

Similar Papers

CASHomon Sets: Efficient Rashomon Sets Across Multiple Model Classes and their Hyperparameters

82% match2026

Exploration of the Rashomon Set Assists Trustworthy Explanations for Medical Data

80% match1 repo12023

Rashomon Sets for Prototypical-Part Networks: Editing Interpretable Models in Real-Time

80% match2025

A Typology for Exploring the Mitigation of Shortcut Behavior

80% match1 repo82022

Archetypal SAE: Adaptive and Stable Dictionary Learning for Concept Extraction in Large Vision Models

80% match2025

Bayesian Inference of Minimally Complex Models with Interactions of Arbitrary Order

80% match1 repo32020

Cite this paper

@article{ross2018learning,
  title  = {Learning Qualitatively Diverse and Interpretable Rules for Classification},
  author = {Andrew Slavin Ross and Weiwei Pan and Finale Doshi-Velez},
  year   = {2018},
  eprint = {1806.08716},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/1806.08716}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/1806.08716.svg)](https://codeofpaper.com/paper/1806.08716)

Discussion