Hyperparameter Tuning
Searching for the combination of a model's untrained settings, like depth or learning rate, that scores best on validation, rather than its learned parameters.
Grid search evaluates every combination on a fixed lattice, so its cost multiplies with every knob added and it spends resolution evenly across dimensions whether or not they matter. Random search draws each hyperparameter independently, so a fixed budget still explores the dimension that actually moves the score at high resolution.
Successive halving cuts the budget a different way: run many candidates cheaply, keep the best fraction, and give only survivors a full training run.