SVEBI: Towards the Interpretation and Explanation of Spiking Neural Networks
Abstract
Artificial Neural Networks have demonstrated the versatility to adapt to different problems while at the same time achieving high predictive performance. A characteristic of these models is their high-computation demands, which translates into high energy consumption. Spiking Neural Networks (SNN) have been proposed as an energy-efficient alternative with promising results in several tasks. Despite being a promising alternative that could motivate massive deployment, research on the interpretation of these models is almost non-existent. To address this problem, we propose SVEBI, a method that extracts insights on the representation encoded by SNNs, via the analysis of sparse relevant internal units that drive the decision-making process. In addition, we show the use of the relevant units as a means to justify, i.e. explain, the predictions made by an SNN for a given input. In this explanation/justification task, our experiments show SVEBI is comparable, if not superior, to existing methods.
Cite
Text
De Laet et al. "SVEBI: Towards the Interpretation and Explanation of Spiking Neural Networks." European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, 2025. doi:10.1007/978-3-032-06066-2_27Markdown
[De Laet et al. "SVEBI: Towards the Interpretation and Explanation of Spiking Neural Networks." European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, 2025.](https://mlanthology.org/ecmlpkdd/2025/laet2025ecmlpkdd-svebi/) doi:10.1007/978-3-032-06066-2_27BibTeX
@inproceedings{laet2025ecmlpkdd-svebi,
title = {{SVEBI: Towards the Interpretation and Explanation of Spiking Neural Networks}},
author = {De Laet, Jasper and Behzadi-Khormouji, Hamed and Deckers, Lucas and Oramas, José},
booktitle = {European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases},
year = {2025},
pages = {460-477},
doi = {10.1007/978-3-032-06066-2_27},
url = {https://mlanthology.org/ecmlpkdd/2025/laet2025ecmlpkdd-svebi/}
}