Happy to share that the paper RAxSS: Retrieval-Augmented Sparse Sampling for Explainable Variable-Length Medical Time Series Classification has been accepted to the Learning from Time Series for Health (TS4H) workshop at NeurIPS 2025 as a joint effort by Aydin Javadov and co-authors.
They introduce a lightweight, retrieval-augmented convex aggregation approach for clinical time series. Evaluated on intracranial EEG data, it shows strong potential for explainable and robust clinical variable-length signal classification.
Preprint available here: https://arxiv.org/pdf/2510.02936
Thanks to co-authors Samir Garibov, Qiyang Sun, Tobias Hoesli, Prof. Dr. Florian Wangenheim, Dr. Joseph Ollier, and Prof. Dr. Björn Schuller
