Epok
The educational unit, which means that the entire educational data is once from the model.
Epok (epoch) is a basic educational unit that refers to a full cycle, based on these examples, all examples of a machine learning model in the educational data set, once from start to finish. The set of data in practice is often processed by split an epok into small groups (batch or mini-batch), since it is so big that it will not fit the memory at once; an update step is thrown when each batch is processed, and an epok is finished when all batches are completed. It is one of the most important hyperparameters of the educational process, which will be trained throughout the few epok of a model.
The models are generally not able to learn enough on a single epokta, so they are gradually improved by passing tens or hundreds of times over the same data. But there is also a price to increase the number of epok: a lot of epok can lead to memorizing the training data of the model, i.e. extreme learning. Therefore, in practice, my success in the verification data set is monitored as soon as the epoxes are progressed and my success is no longer recovered or the training is stopped early (early stopping). The concept of epok is used as a standard measure unit in almost all surveillance and unsupervised deep learning training processes from major language models to image classifiers.
The models are generally not able to learn enough on a single epokta, so they are gradually improved by passing tens or hundreds of times over the same data. But there is also a price to increase the number of epok: a lot of epok can lead to memorizing the training data of the model, i.e. extreme learning. Therefore, in practice, my success in the verification data set is monitored as soon as the epoxes are progressed and my success is no longer recovered or the training is stopped early (early stopping). The concept of epok is used as a standard measure unit in almost all surveillance and unsupervised deep learning training processes from major language models to image classifiers.