A hyperparameter is a configuration value set before or during model training that controls how learning occurs.
Examples include learning rate, batch size, number of layers, regularization settings, and other choices tuned to improve model performance.