Overfitting
/ˌoʊvərˈfɪtɪŋ/
When a model learns training data too well, including noise, resulting in poor generalization to new data.
In italiano: OverfittingOverfitting occurs when models are too complex relative to training data size. Solutions include regularization, dropout, early stopping, and data augmentation. Balance between underfitting and overfitting is crucial.
Examples
- A decision tree memorizing all training examples
- Neural network with too many parameters
- Model performing 100% on training but 60% on test data