New Member Double-Header

Zoom information will be sent to e-HAIL members.

Data Quality and Standards for AI and Learning Health Systems
Data standards are essential infrastructure for learning health systems and underpin data quality and readiness for AI. I will highlight some relevant standards —including those for data provenance, longitudinal health records, and clinical decision support—that will support the generation and application of AI models at scale and highlight the role of these standards in rapidly evolving health data infrastructure.

Interpretable Learning Methods for Biomedical Data: A “Choose Your Own Adventure” Talk
Biomedical data is often plagued by small sample sizes, large feature sets, and high noise. The lasso is a popular prediction method that addresses these issues by selecting the subset of features that are most useful for prediction. We will introduce the lasso and invite the audience to guide the remainder of the talk, covering either (1) a method for lasso pretraining, which allows us to use the lasso to leverage shared signal across datasets (e.g. across cancer types) or (2) a method that uses the lasso to identify disease-associated cells in single-cell data without cell-level annotations.