An Epoch is one complete pass through the entire training dataset by a machine learning model. Multiple epochs are typically required for a model to converge. The choice of how many epochs to train can influence both performance and risk of overfitting.