Overfitting

Overfitting occurs when a model fits its training data too closely, learning random noise and coincidences instead of the underlying pattern. It then fails to generalize to new data.

Why it matters

An overfit model looks excellent in development and disappoints in production. Catching it requires evaluating on data the model never trained on. The opposite problem, underfitting, is a model too simple to capture the pattern at all. It is one of the most common reasons a promising prototype fails after launch.

Example

Typical symptom:

training accuracy:   99%
validation accuracy: 71%

The large gap signals memorization. Remedies include more varied data, simpler models, regularization, dropout, early stopping, and cross-validation.

Common mistakes

• Tuning settings repeatedly against the test set, which quietly overfits to it.
• Reading a high training score as proof the model works.
• Fixing it only by adding complexity, which often makes it worse.

Related

• Machine learning
• Supervised learning
• Dataset
• Neural network

This page is general information written by our editorial team and updated when documentation changes. Servers and providers differ, so test changes safely first. See our editorial policy and disclaimer.