A Consistent and Efficient Evaluation Strategy for Attribution Methods

Yao Rong, Tobias Leemann, Vadim Borisov, Gjergji Kasneci, Enkelejda Kasneci
Proceedings of the 39th International Conference on Machine Learning, PMLR 162:18770-18795, 2022.

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.

Cite this Paper

BibTeX
@InProceedings{pmlr-v162-rong22a, title = {A Consistent and Efficient Evaluation Strategy for Attribution Methods}, author = {Rong, Yao and Leemann, Tobias and Borisov, Vadim and Kasneci, Gjergji and Kasneci, Enkelejda}, booktitle = {Proceedings of the 39th International Conference on Machine Learning}, pages = {18770--18795}, year = {2022}, editor = {Chaudhuri, Kamalika and Jegelka, Stefanie and Song, Le and Szepesvari, Csaba and Niu, Gang and Sabato, Sivan}, volume = {162}, series = {Proceedings of Machine Learning Research}, month = {17--23 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v162/rong22a/rong22a.pdf}, url = {https://proceedings.mlr.press/v162/rong22a.html}, 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.} }
Endnote
%0 Conference Paper %T A Consistent and Efficient Evaluation Strategy for Attribution Methods %A Yao Rong %A Tobias Leemann %A Vadim Borisov %A Gjergji Kasneci %A Enkelejda Kasneci %B Proceedings of the 39th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2022 %E Kamalika Chaudhuri %E Stefanie Jegelka %E Le Song %E Csaba Szepesvari %E Gang Niu %E Sivan Sabato %F pmlr-v162-rong22a %I PMLR %P 18770--18795 %U https://proceedings.mlr.press/v162/rong22a.html %V 162 %X 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.
APA
Rong, Y., Leemann, T., Borisov, V., Kasneci, G. & Kasneci, E.. (2022). A Consistent and Efficient Evaluation Strategy for Attribution Methods. Proceedings of the 39th International Conference on Machine Learning, in Proceedings of Machine Learning Research 162:18770-18795 Available from https://proceedings.mlr.press/v162/rong22a.html.

Related Material