TAMER: A Test-Time Adaptive MoE-Driven Framework for EHR Representation Learning
Zhu, Yinghao, Zheng, Xiaochen, Allam, Ahmed, Krauthammer, Michael
–arXiv.org Artificial Intelligence
We propose TAMER, a Test-time Adaptive MoE-driven framework for EHR Representation learning. TAMER combines a Mixture-of-Experts (MoE) with Test-Time Adaptation (TTA) to address two critical challenges in EHR modeling: patient population heterogeneity and distribution shifts. The MoE component handles diverse patient subgroups, while TTA enables real-time adaptation to evolving health status distributions when new patient samples are introduced. Extensive experiments across four real-world EHR datasets demonstrate that TAMER consistently improves predictive performance for both mortality and readmission risk tasks when combined with diverse EHR modeling backbones. TAMER offers a promising approach for dynamic and personalized EHR-based predictions in practical clinical settings. Code is publicly available at https://github.com/yhzhu99/TAMER.
arXiv.org Artificial Intelligence
Jan-9-2025
- Country:
- Europe > Switzerland > Zürich > Zürich (0.14)
- Genre:
- Research Report
- Experimental Study (0.68)
- New Finding (0.46)
- Research Report
- Industry:
- Health & Medicine > Therapeutic Area (1.00)
- Technology: