We discuss a way to find a well behaved baseline for attribution methods that work by feeding a neural network with a sequence of interpolated inputs between two given inputs. Then, we test it with our novel Riemann-Stieltjes Integrated Gradient-weighted Class Activation Mapping (RSI-Grad-CAM) attribution method.
Cite
@article{arxiv.2204.06120,
title = {Baseline Computation for Attribution Methods Based on Interpolated Inputs},
author = {Miguel Lerma and Mirtha Lucas},
journal= {arXiv preprint arXiv:2204.06120},
year = {2022}
}