BEE: Metric-Adapted Explanations via Baseline Exploration-Exploitation

TL;DR

This paper introduces BEE, a novel method that adaptively generates explanations by modeling baselines as learned distributions, improving explanation quality across diverse metrics through a unified exploration-exploitation approach.

Contribution

BEE provides a unified framework for explanation evaluation by modeling baselines as learned distributions and optimizing explanations for specific metrics via exploration-exploitation.

Findings

BEE outperforms state-of-the-art explanation methods across various metrics.

Modeling baselines as learned distributions enhances explanation adaptability.

Extensive evaluations demonstrate BEE's superior performance.

Abstract

Two prominent challenges in explainability research involve 1) the nuanced evaluation of explanations and 2) the modeling of missing information through baseline representations. The existing literature introduces diverse evaluation metrics, each scrutinizing the quality of explanations through distinct lenses. Additionally, various baseline representations have been proposed, each modeling the notion of missingness differently. Yet, a consensus on the ultimate evaluation metric and baseline representation remains elusive. This work acknowledges the diversity in explanation metrics and baselines, demonstrating that different metrics exhibit preferences for distinct explanation maps resulting from the utilization of different baseline representations and distributions. To address the diversity in metrics and accommodate the variety of baseline representations in a unified manner, we propose Baseline Exploration-Exploitation (BEE) - a path-integration method that introduces randomness to the integration process by modeling the baseline as a learned random tensor. This tensor follows a learned mixture of baseline distributions optimized through a contextual exploration-exploitation procedure to enhance performance on the specific metric of interest. By resampling the baseline from the learned distribution, BEE generates a comprehensive set of explanation maps, facilitating the selection of the best-performing explanation map in this broad set for the given metric. Extensive evaluations across various model architectures showcase the superior performance of BEE in comparison to state-of-the-art explanation methods on a variety of objective evaluation metrics.