Hyperparameter — A hyperparameter is a setting chosen by people before training, rather than learned by the model. Examples include the learning rate, batch size, and number of layers. Good hyperparameters can make a large difference in results. Finding them often takes many experiments.

Added October 21, 2026