All terms
Foundations
Generalization
How well a model applies what it learned to fresh, unseen inputs.
Definition
Generalization is how well a model applies what it learned to fresh, unseen inputs rather than just the examples it trained on. Strong generalization is the real goal of machine learning, since a model that only memorizes its training data is of little use. It is the opposite failure mode from overfitting, and it is measured on held-out data the model did not train on.