Hyperparameter

A configuration choice that governs a model or its training, rather than being directly learned from the training data. Evaluate choices using suitable validation and record them for reproducibility.