Explaining the Explainer: A First Theoretical Analysis of LIME

TL;DR

This paper provides the first theoretical analysis of LIME, showing it accurately captures meaningful features when parameters are well-chosen, but can miss important features otherwise.

Contribution

It derives closed-form expressions for LIME's coefficients in linear cases and analyzes how parameter choices affect feature detection.

Findings

LIME coefficients are proportional to the gradient of the function.

Poor parameter choices can cause LIME to miss important features.

LIME effectively discovers meaningful features with proper parameter tuning.

Abstract

Machine learning is used more and more often for sensitive applications, sometimes replacing humans in critical decision-making processes. As such, interpretability of these algorithms is a pressing need. One popular algorithm to provide interpretability is LIME (Local Interpretable Model-Agnostic Explanation). In this paper, we provide the first theoretical analysis of LIME. We derive closed-form expressions for the coefficients of the interpretable model when the function to explain is linear. The good news is that these coefficients are proportional to the gradient of the function to explain: LIME indeed discovers meaningful features. However, our analysis also reveals that poor choices of parameters can lead LIME to miss important features.