Configuration settings that control the learning process of machine learning algorithms, set before training begins. Examples include learning rate, batch size, and number of hidden layers.