Hyperparameter
In plain terms
A configuration value set before training, not learned from data.
Going deeper
Hyperparameters (like learning rate or tree depth) control the training process and model structure, tuned to optimize performance. They shape how the model learns.
Under the hood
Tuned via grid/random search or Bayesian optimization; distinct from parameters learned during training.
Examples
Learning rate, batch size, number of layers, tuning, grid search.
Think of it like...
Like the oven temperature and time you set before baking, not the cake itself.