2017

Interpretable & Explorable Approximations of Black Box Models

Lakkaraju, Himabindu, Kamar, Ece, Caruana, Rich et al.

Understand

We propose Black Box Explanations through Transparent Approximations (BETA), a novel model agnostic framework for explaining the behavior of any black-box classifier by simultaneously optimizing for fidelity to the original model and interpretability of the explanation.

  • To this end, we develop a novel objective function which allows us to learn (with optimality guarantees), a small number of compact decision sets each of which explains the behavior of the black box model in unambiguous, well-defined regions of feature space.
  • Furthermore, our framework also is capable of accepting user input when generating these approximations, thus allowing users to interactively explore how the black-box model behaves in different subspaces that are of interest to the user.
  • To the best of our knowledge, this is the first approach which can produce global explanations of the behavior of any given black box model through joint optimization of unambiguity, fidelity, and interpretability, while also allowing users to explore model behavior based on their preferences.

Built on

  • The budgeted maximum coverage problem

    S. Khuller, A. Moss, and J. S. Naor · 1999

    Earlier work this paper cites.

  • Top-down induction of decision trees classifiers-a survey

    L. Rokach and O. Maimon · 2005

    Earlier work this paper cites.

  • Non-monotone submodular maximization under matroid and knapsack constraints

    J. Lee, V. S. Mirrokni, V. Nagarajan, and M. Sviridenko · 2009

    Earlier work this paper cites.

Similar

  • Intelligible models for classification and regression

    Y. Lou, R. Caruana, and J. Gehrke · 2012

    Cited alongside, same era.

  • Interpretable classifiers using rules and bayesian analysis: Building a better stroke prediction model

    B. Letham, C. Rudin, T. H. McCormick, D. Madigan, et al · 2015

    Cited alongside, same era.

  • Interpretable decision sets: A joint framework for description and prediction

    H. Lakkaraju, S. H. Bach, and J. Leskovec · 2016

    Cited alongside, same era.

  • Fast algorithms for mining association rules

    R. Agrawal, R. Srikant, et al

    Cited in the paper.

Then

Beyond the bibliography

alphaXiv searches the wider corpus for related work and actual follow-ups.

Open on alphaXiv

alphaXiv is searching for related work…