Data-Efficient EHR Learning

Background: Learning from finite, longitudinal electronic health records (EHRs) requires algorithms that maximize data efficiency, contrasting with language models trained on web-scale corpora where data is near-infinite.

Question / Future Work: Developing more data-efficient algorithms for learning robust representations from longitudinal EHRs is crucial, especially considering the limited and often scarce nature of patient data. This includes investigating methods for generating high-quality synthetic data to augment real, constrained data sources.