Enhancing Spectral Embedding through Robust and Flexible Knowledge Transfer in Electronic Health Records
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 method departs from existing approaches by relaxing restrictive one-to-one signal-alignment assumptions between the latent data matrix and knowledge matrix, allowing more flexible and realistic forms of structured sharing. We introduce a novel two-step spectral embedding procedure: first, we identify and remove irrelevant components from the knowledge matrix; then, we apply a projection-based method to separately recover shared and heterogeneous components. Simulations and an analysis of a real-world multiple sclerosis cohort show that the proposed method outperforms competing approaches, particularly in challenging scenarios where shared signals are weak and only partially aligned, as is common in rare-disease data.
Related
- Spectral embedding through weak* limit of finite-dimensional approximations
- Group Invariant Spectral Embedding
- Null-Space-Free 6D Spectral Embedding with Local Rayleigh Quotient Recovery for 3D Quasiperiodic Maxwell's Eigenproblems
- ISOMETRIC EMBEDDING OF NONCOMMUTATIVE L_p -SPACES INTO NONCOMMUTATIVE SYMMETRIC SPACES
- Embedding theorems for Herz-Type Sobolev spaces on domains
BibTeX
@misc{huang2026enhancing,
title = {{Enhancing Spectral Embedding through Robust and Flexible Knowledge Transfer in Electronic Health Records}},
author = {Feiqing Huang and Zongqi Xia and Rong Ma and Tianxi Cai},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2606.11570},
eprint = {2606.11570},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2606.11570},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Enhancing Spectral Embedding through Robust and Flexible Knowledge Transfer in Electronic Health Records
AU - Feiqing Huang
AU - Zongqi Xia
AU - Rong Ma
AU - Tianxi Cai
PY - 2026
DA - 2026/10/05
PB - arXiv
DO - 10.48550/arxiv.2606.11570
UR - https://arxiv.org/abs/2606.11570
AB - 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 method departs from existing approaches by relaxing restrictive one-to-one signal-alignment assumptions between the latent data matrix and knowledge matrix, allowing more flexible and realistic forms of structured sharing. We introduce a novel two-step spectral embedding procedure: first, we identify and remove irrelevant components from the knowledge matrix; then, we apply a projection-based method to separately recover shared and heterogeneous components. Simulations and an analysis of a real-world multiple sclerosis cohort show that the proposed method outperforms competing approaches, particularly in challenging scenarios where shared signals are weak and only partially aligned, as is common in rare-disease data.
ER -