ProtoKDM: An interpretable prototype-based framework for biomedical time-series analysis

doi:10.1016/j.bspc.2026.109473

ProtoKDM: An interpretable prototype-based framework for biomedical time-series analysis

2026-01-07

0

OA

AI

DOI：10.1016/j.bspc.2026.109473

Original

Save

View PDF

Abstract

En 中文

• ProtoKDM explains biomedical time-series using density-aware prototypes. • Bifurcated autoencoder enhances class separation in latent representation. • Kernel Density Matrices enable clinically grounded prototype selection. • UMAP-based visualizations support global interpretability. • Signal-level comparisons enhance transparency and decision support.

Keywords:

Explainable AI

Prototype learning

Latent-space representations

Biomedical signals

Deep learning

ProtoKDM: An interpretable prototype-based framework for biomedical time-series analysis

Abstract

AI Summary

Journal

Organization