Anomalous behaviour in loss-gradient based interpretability methods
Machine Learning
2022-07-19 v1 Artificial Intelligence
Information Retrieval
Abstract
Loss-gradients are used to interpret the decision making process of deep learning models. In this work, we evaluate loss-gradient based attribution methods by occluding parts of the input and comparing the performance of the occluded input to the original input. We observe that the occluded input has better performance than the original across the test dataset under certain conditions. Similar behaviour is observed in sound and image recognition tasks. We explore different loss-gradient attribution methods, occlusion levels and replacement values to explain the phenomenon of performance improvement under occlusion.
Cite
@article{arxiv.2207.07769,
title = {Anomalous behaviour in loss-gradient based interpretability methods},
author = {Vinod Subramanian and Siddharth Gururani and Emmanouil Benetos and Mark Sandler},
journal= {arXiv preprint arXiv:2207.07769},
year = {2022}
}
Comments
Accepted at ICLR RobustML workshop 2021