A Consistent and Efficient Evaluation Strategy for Attribution Methods
arXiv:2202.00449
Abstract
With a variety of local feature attribution methods being proposed in recent years, follow-up work suggested several evaluation strategies. To assess the attribution quality across different attribution techniques, the most popular among these evaluation strategies in the image domain use pixel perturbations. However, recent advances discovered that different evaluation strategies produce conflicting rankings of attribution methods and can be prohibitively expensive to compute. In this work, we present an information-theoretic analysis of evaluation strategies based on pixel perturbations. Our findings reveal that the results are strongly affected by information leakage through the shape of the removed pixels as opposed to their actual values. Using our theoretical insights, we propose a novel evaluation framework termed Remove and Debias (ROAD) which offers two contributions: First, it mitigates the impact of the confounders, which entails higher consistency among evaluation strategies. Second, ROAD does not require the computationally expensive retraining step and saves up to 99% in computational costs compared to the state-of-the-art. We release our source code at https://github.com/tleemann/road_evaluation.
26 pages. Accepted at ICML 2022
Cited by in corpus (6)
- Towards Human-centered Explainable AI: A Survey of User Studies for Model Explanations
- Generalizable and Explainable Deep Learning for Medical Image Computing: An Overview
- T-TAME: Trainable Attention Mechanism for Explaining Convolutional Networks and Vision Transformers
- Analyzing Effects of Mixed Sample Data Augmentation on Model Interpretability
- On Spectral Properties of Gradient-based Explanation Methods
- Don't be fooled: label leakage in explanation methods and the importance of their quantitative evaluation