Learning with Missing Values: Theoretical Insights and Application to Health Databases

Explore the challenges and solutions for handling missing values in supervised learning tasks through this insightful lecture. Delve into the theoretical foundations of Impute-then-Regress approaches and their practical applications in health databases. Discover how different baseline methods for handling missing values compare across large health datasets with naturally occurring missing data. Gain valuable insights into regression and classification tasks in the presence of incomplete information. Learn about a novel neural network architecture designed specifically for learning with missing values, surpassing traditional two-stage Impute-then-Regress approaches. Understand the importance of addressing missing data in fields such as health, business, and social sciences, and acquire knowledge on cutting-edge techniques to improve data analysis and decision-making processes.