On the Art and Science of Machine Learning Explanations

TL;DR

This paper reviews various machine learning explanation methods, discussing their theoretical foundations, practical applications, and providing usage recommendations with software examples to enhance interpretability.

Contribution

It offers a comprehensive overview of popular explanation techniques, comparing their scope, fidelity, and suitability, supported by real-world use cases and reproducible software resources.

Findings

Decision tree surrogate models are useful for interpretability.

LIME and Shapley explanations provide local interpretability.

Partial dependence plots help understand feature effects.

Abstract

This text discusses several popular explanatory methods that go beyond the error measurements and plots traditionally used to assess machine learning models. Some of the explanatory methods are accepted tools of the trade while others are rigorously derived and backed by long-standing theory. The methods, decision tree surrogate models, individual conditional expectation (ICE) plots, local interpretable model-agnostic explanations (LIME), partial dependence plots, and Shapley explanations, vary in terms of scope, fidelity, and suitable application domain. Along with descriptions of these methods, this text presents real-world usage recommendations supported by a use case and public, in-depth software examples for reproducibility.