Sampling Based On Natural Image Statistics Improves Local Surrogate Explainers
Ricardo Kleinlein (Universidad Politécnica de Madrid),* Alexander Hepburn (University of Bristol), Raul Santos Rodriguez (University of Bristol), Fernando Fernández-Martínez (Universidad Politécnica de Madrid)
The 33rd British Machine Vision Conference
Abstract
Many problems in computer vision are recently been tackled using deep neural networks, whose predictions cannot be easily interpreted. Surrogate explainers aim to address this, as a popular post-hoc interpretability method to further understand how a black-box model arrives at a particular prediction. By training a simple, more interpretable model to locally approximate the decision boundary of a non-interpretable system, we can estimate the relative importance of the input features on the prediction. Focusing on images, most surrogate explainers, e.g., LIME, generate a local neighbourhood around a query image by sampling in an interpretable domain. However, interpretable domains have traditionally been derived exclusively from the intrinsic features of the query image, not taking into consideration the manifold of the data the non-interpretable model has been exposed to in training (or more generally, the manifold of real images). This leads to suboptimal surrogates as they are trained on images that lie within low probability regions of the manifold of real images. In this work, we address this limitation by aligning the local neighbourhood on which the surrogate is trained with the original training data distribution, even when this distribution is not accessible. We propose two approaches to do so, namely (1) altering the method for sampling the local neighbourhood and (2) using perceptual metrics to convey some of the properties of the statistics of natural images.Video
Citation
@inproceedings{Kleinlein_2022_BMVC,
author = {Ricardo Kleinlein and Alexander Hepburn and Raul Santos Rodriguez and Fernando Fernández-Martínez },
title = {Sampling Based On Natural Image Statistics Improves Local Surrogate Explainers},
booktitle = {33rd British Machine Vision Conference 2022, {BMVC} 2022, London, UK, November 21-24, 2022},
publisher = {{BMVA} Press},
year = {2022},
url = {https://bmvc2022.mpi-inf.mpg.de/1083.pdf}
}
Copyright © 2022
The British Machine Vision Conference is organised by