What is early stopping and how does it prevent overfitting?

Early stopping halts training when validation performance stops improving.

During iterative training, the model is evaluated on validation data. If performance does not improve for a defined number of iterations, training stops, preventing overfitting to training data noise.