• Computer Science
  • Published in NIPS 2016

Examples are not enough, learn to criticize! Criticism for Interpretability

@inproceedings{Kim2016ExamplesAN,
  title={Examples are not enough, learn to criticize! Criticism for Interpretability},
  author={Been Kim and Oluwasanmi Koyejo and Rajiv Khanna},
  booktitle={NIPS},
  year={2016}
}
Example-based explanations are widely used in the effort to improve the interpretability of highly complex distributions. However, prototypes alone are rarely sufficient to represent the gist of the complexity. In order for users to construct better mental models and understand complex data distributions, we also need {\em criticism} to explain what are \textit{not} captured by prototypes. Motivated by the Bayesian model criticism framework, we develop \texttt{MMD-critic} which efficiently… CONTINUE READING

Figures and Topics from this paper.

Citations

Publications citing this paper.
SHOWING 1-10 OF 106 CITATIONS

Comparative Document Summarisation via Classification

VIEW 10 EXCERPTS
CITES BACKGROUND, RESULTS & METHODS
HIGHLY INFLUENCED

MARGIN: Uncovering Deep Neural Networks using Graph Signal Analysis

VIEW 8 EXCERPTS
CITES METHODS, RESULTS & BACKGROUND
HIGHLY INFLUENCED

Efficient Data Representation by Selecting Prototypes with Importance Weights

VIEW 21 EXCERPTS
CITES BACKGROUND & METHODS
HIGHLY INFLUENCED

ConvNets and ImageNet Beyond Accuracy: Understanding Mistakes and Uncovering Biases

VIEW 4 EXCERPTS
CITES METHODS
HIGHLY INFLUENCED

Influential Sample Selection: A Graph Signal Processing Approach

VIEW 10 EXCERPTS
CITES RESULTS, METHODS & BACKGROUND
HIGHLY INFLUENCED

ProtoDash: Fast Interpretable Prototype Selection

VIEW 10 EXCERPTS
CITES BACKGROUND & METHODS
HIGHLY INFLUENCED

Explainability fact sheets: a framework for systematic assessment of explainable approaches

VIEW 4 EXCERPTS
CITES BACKGROUND
HIGHLY INFLUENCED

On Human Predictions with Explanations and Predictions of Machine Learning Models: A Case Study on Deception Detection

VIEW 4 EXCERPTS
CITES METHODS & BACKGROUND
HIGHLY INFLUENCED

FILTER CITATIONS BY YEAR

2016
2020

CITATION STATISTICS

  • 12 Highly Influenced Citations

  • Averaged 33 Citations per year from 2017 through 2019

  • 80% Increase in citations per year in 2019 over 2018

References

Publications referenced by this paper.
SHOWING 1-10 OF 29 REFERENCES

An analysis of approximations for maximizing submodular set functions—I

VIEW 4 EXCERPTS
HIGHLY INFLUENTIAL

A Class of Submodular Functions for Document Summarization

VIEW 4 EXCERPTS
HIGHLY INFLUENTIAL

A Database for Handwritten Text Recognition Research

VIEW 3 EXCERPTS
HIGHLY INFLUENTIAL

Engineering safety in machine learning

  • Kush R. Varshney
  • Computer Science, Mathematics
  • 2016 Information Theory and Applications Workshop (ITA)
  • 2016
VIEW 1 EXCERPT

Deep Residual Learning for Image Recognition

VIEW 1 EXCERPT

On Greedy Maximization of Entropy

VIEW 1 EXCERPT