Cost is another term for a model's loss or objective value during training or evaluation. It represents the penalty associated with prediction error and is typically minimized by the learning algorithm.
Synonym for loss.