Epoch
One complete pass through the entire training dataset during the training process of a machine learning model.
What is an Epoch?
During model training, data is typically divided into smaller groups called batches. When the model has processed every batch in the training dataset once, it has completed one epoch. Models are usually trained for multiple epochs, allowing their parameters to be repeatedly updated as they learn patterns and reduce prediction errors.
Why is an Epoch Important?
The number of epochs affects how well a model learns from its training data. Too few epochs can result in underfitting, while too many may cause overfitting. Monitoring training and validation performance across epochs helps teams determine when a model has learned sufficiently and when training should stop.
Common use cases
Epochs are commonly used when training neural networks, deep learning models, image classifiers, language models, and other models trained through iterative optimization.