Statistical Methods for Learning and Discovery

Catalog description

This course introduces modern machine learning methods with a focus on applications in healthcare data. Students will explore both foundational and advanced topics in supervised and unsupervised learning tailored to the complexities of clinical and multi-modal data. Topics include computable phenotyping, risk prediction, causal inference, model evaluation, and bias mitigation. Emphasis is placed on the unique challenges of working with structured and unstructured health data, including missingness, fairness, and interpretability. Methods covered include regularized regression, decision trees and ensemble models, smoothing and splines, clustering, and dimensionality reduction. Coursework involves hands-on programming assignments and real-world case studies using Python and R. Prior knowledge of regression and statistical modeling and Python programming is expected. Prerequisite(s): Permission only. Credits: 3.