Early stopping: Difference between revisions

Machine learning seeks to train a model that can make accurate predictions on new data. Unfortunately, during training it is common for the model to overfit the training data; that is, it becomes too complex and includes irrelevant details or noise in the dataset. Unfortunately, overfitting can lead to poor performance when faced with new scenarios - thus defeating its purpose. Early stopping is an approach used in machine learning to prevent overfitting and improve generalization by monitoring model performance during training and stopping when its performance on validation sets starts to deteriorate.