Caesar AI Atlas

Empirical Risk Minimization (ERM)

Also known as: ERM

Caesar AI Atlas Definition

ERM (Empirical Risk Minimization) is a learning principle that selects the function or model that minimizes loss on the training data. It is a foundational approach in supervised learning and contrasts with methods that explicitly control model complexity or expected generalization risk.

Other Definitions

Empirical Risk Minimization Source

Choosing the function that minimizes loss on the training set. Contrast with structural risk minimization.

Related Terms