Accelerating the Global Aggregation of Local Explanations

TL;DR

This paper introduces techniques to significantly accelerate the aggregation of local explanations, like the Anchor algorithm, enabling faster global insights into model behavior with minimal loss of accuracy.

Contribution

It proposes lossless and lossy acceleration methods for aggregating local explanations, reducing computation time by up to 30 times while maintaining explanation quality.

Findings

Achieved up to 30× speedup in aggregation process

Reduced computation time from hours to minutes

Developed a probabilistic model to reduce bias in impact estimation

Abstract

Local explanation methods highlight the input tokens that have a considerable impact on the outcome of classifying the document at hand. For example, the Anchor algorithm applies a statistical analysis of the sensitivity of the classifier to changes in the token. Aggregating local explanations over a dataset provides a global explanation of the model. Such aggregation aims to detect words with the most impact, giving valuable insights about the model, like what it has learned in training and which adversarial examples expose its weaknesses. However, standard aggregation methods bear a high computational cost: a na\"ive implementation applies a costly algorithm to each token of each document, and hence, it is infeasible for a simple user running in the scope of a short analysis session. % We devise techniques for accelerating the global aggregation of the Anchor algorithm. Specifically, our goal is to compute a set of top-$k$ words with the highest global impact according to different aggregation functions. Some of our techniques are lossless and some are lossy. We show that for a very mild loss of quality, we are able to accelerate the computation by up to 30$\times$, reducing the computation from hours to minutes. We also devise and study a probabilistic model that accounts for noise in the Anchor algorithm and diminishes the bias toward words that are frequent yet low in impact.