What technique involves stopping the training process as soon as the validation error begins to increase?
Show answer
Early stopping
Early stopping is a form of regularization used to avoid overfitting. By monitoring the performance on a separate validation set, training is halted at the point of best generalization before the model starts learning noise.
Also asked as:
- What regularization technique halts the training of a neural network at the point of best generalization, before the validation error starts to rise?