Early Stopping
Halting training when validation performance stops improving — a simple, effective guard against overfitting.
Early stopping is a regularization technique: you monitor performance on a held-out validation set during training and stop when it stops improving (and starts to worsen), instead of training for a fixed number of epochs. It prevents the model from continuing into the overfitting regime.
Worked example: validation loss falls for 30 epochs then begins to rise while training loss keeps dropping — early stopping halts around epoch 30 (often restoring the best checkpoint), capturing the model before it memorized noise. Gotcha: it needs a “patience” buffer (wait N epochs of no improvement before stopping) because validation loss is noisy and can dip then recover — stopping at the first uptick quits too early; and it relies on a validation set that truly represents production, or you stop on the wrong signal.