XRAI: Better Attributions Through Regions
2019pp. 4947–4956
Citations Over TimeTop 10% of 2019 papers
Abstract
Saliency methods can aid understanding of deep neural networks. Recent years have witnessed many improvements to saliency methods, as well as new ways for evaluating them. In this paper, we 1) present a novel region-based attribution method, XRAI, that builds upon integrated gradients (Sundararajan et al. 2017), 2) introduce evaluation methods for empirically assessing the quality of image-based saliency maps (Performance Information Curves (PICs)), and 3) contribute an axiom-based sanity check for attribution methods. Through empirical experiments and example results, we show that XRAI produces better results than other saliency methods for common models and the ImageNet dataset.
Related Papers
- → Between Sanity and Madness(2019)1 cited
- → Talk Given at the "Prophet of Sanity" Conference, 26-28 June 1997(1997)
- On the Construction of Undergraduates' Sanity Personality under Globalized Background(2005)
- → Scheme of human development (inspired by Moreno and Buber): an approach to sanity and insanity(2004)
- → THE ISSUES OF ESTABLISHING GUILT IN CASE OF RECOGNIZING A PERSON TO BE INSANE(2023)