Сходимость — это состояние обучения, при котором loss модели или другая optimization objective изменяется очень мало между итерациями. Она указывает, что процесс оптимизации приблизился к стабильному решению, хотя не обязательно к наилучшему возможному.
A state reached when loss values change very little or not at all with each iteration. For example, the following loss curve suggests convergence at around 700 iterations: !Cartesian plot. X-axis is loss. Y-axis is the number of training iterations. Loss is very high during first few iterations, but drops sharply. After about 100 iterations, loss is still descending but far more gradually. After about 700 iterations, loss stays flat. A model converges when additional training won't improve the model. In deep learning, loss values sometimes stay constant or nearly so for many iterations before finally descending. During a long period of constant loss values, you may temporarily get a false sense of convergence. See also early stopping. See Model convergence and loss curves in Machine Learning Crash Course for more information.