Enhancing Spectral Embedding through Robust and Flexible Knowledge Transfer in Electronic Health Records

Excerpt

arXiv:2606.11570v2 Announce Type: replace-cross Abstract: We propose a spectral-based, unsupervised representation learning framework to derive low-dimensional embeddings for clinical concepts and patients in rare disease cohorts from electronic health records, where data are high-dimensional but sample sizes are limited. To overcome this challenge, we incorporate a knowledge matrix extracted from a broader population that shares a partially overlapping subspace with the rare-disease cohort. Our