Generalization
Domain: Evaluation
🎯 Motivation
The model’s ability to perform well on new, unseen data. The ultimate goal of training.
📋 Concrete Example
Model trained on Hong Kong weather generalizes well to Tokyo weather (similar patterns). But fails on Sahara weather (very different). Generalization depends on training data diversity.
🔗 Analogy
A student who understands concepts, not just memorizes - they can solve new problems they have never seen. Generalization is applying learned patterns to new situations.