Generalization Is the Real Test
Explain why neural-network performance must be checked on validation or test data, not only training data.
Training performance is practice; unseen-data performance is the test. Neural networks can fit training examples very well, especially when they have many parameters. That is useful only if the learned pattern transfers to new cases. Generalization is the model's ability to perform on data it did not use to learn the weights. This is why teams split data into training, validation, and test sets or use cross-validation. Training data teaches the model. Validation data helps tune choices such as architecture, learning rate, or regularization. Test data provides the more honest final estimate because it is held back from the learning…
Sign up free — one personalized lesson every day, matched to your role and goals.
Already have an account? Sign in