Machine Learning

Generalization

A model's ability to perform well on new, unseen data that was not part of its training set. Generalization is the ultimate goal of machine learning — learning patterns, not memorizing examples.

Why It Matters

A model that does not generalize is useless in production. Generalization is what separates a useful AI system from an expensive lookup table.

Example

A spam classifier trained on 2024 emails still correctly identifying spam in 2025 despite evolving spam tactics — it learned the underlying patterns, not specific examples.

Think of it like...

Like learning to drive in one city and being able to drive competently in any city — you learned the principles of driving, not just one specific route.

Related Terms