← Publications · 2026 · conference
Saheed Faremi · Andrea Visentin · Luca Longo
XAI 2026 (Late-breaking work + Doctoral Consortium track), Fortaleza, Brazil. arXiv preprint.
Abstract
A Convolutional Variational Deep Embedding (Conv-VaDE) model for analysing brain electrical activity. The approach jointly learns topographic reconstruction and probabilistic soft clustering in a shared latent space, replacing traditional hard-assignment methods with probabilistic alternatives. Through a systematic architecture search across cluster counts (K=3-20), latent dimensionality, network depth, and channel width, optimal results are reported at depth L=4, achieving a best-case GEV of 0.730 and a silhouette of 0.229 at K=4. Tested on the LEMON resting-state EEG dataset.
Co-located with the 4th World Conference on eXplainable Artificial Intelligence (XAI 2026), Fortaleza, Brazil, 1-3 July 2026. School of Computer Science and IT, University College Cork. Affiliated with the AI and Cognitive Load Research Lab (UCC) and the Insight SFI Research Centre for Data Analytics (UCC).
Companion repo with the full sweep code and 4,832 trained-model results: microstate-architecture-search. A stage-by-stage walkthrough of the preprocessing pipeline is on the blog.
@inproceedings{faremi2026interpretable,
author = {Saheed Faremi and Andrea Visentin and Luca Longo},
title = {Interpretable EEG Microstate Discovery via Variational Deep Embedding: A Systematic Architecture Search with Multi-Quadrant Evaluation},
booktitle = {XAI 2026 (Late-breaking work + Doctoral Consortium track), Fortaleza, Brazil. arXiv preprint.},
year = {2026},
url = {https://arxiv.org/abs/2605.10947}
}