On the explainability of max-plus neural networks
Abstract
We investigate the explanability properties of the recently proposed linear-min-max neural networks. At initialization, they can be interpreted as k-medoids with the infinity norm as a distance. Then, they are trained using subgradient descent to better fit the data. The model has been shown to be a universal approximator. Yet, we can trace the decision process because a single most activated neuron is responsible for the value of the output. Using this property, we designed a pixel fragility measure that determines whether changes to a single pixel may be responsible to a change in the classification output. Experiments on the PneumoniaMnist dataset show that this explanation for the output of the neural network compares favorably to SHAP and Integrated Gradient.
Cite
@article{arxiv.2605.00889,
title = {On the explainability of max-plus neural networks},
author = {Ikhlas Enaieh and Olivier Fercoq and García Ángel},
journal= {arXiv preprint arXiv:2605.00889},
year = {2026}
}
Comments
IEEE International Symposium on Computer-Based Medical Systems (CBMS 2026), Jun 2026, Limassol, Cyprus, Cyprus