The algorithm you chose yesterday has knobs like max_depth, eta, num_round. These hyperparameters are values set before training and set wrongly, even a good algorithm performs poorly. The problem is there are dozens or hundreds of combinations—testing each manually is impractical. SageMaker's Automatic Model Tuning (AMT), aka Hyperparameter Tuning, automates this search.
In the MLA-C01 exam, AMT appears as keywords like "search strategy (Bayesian/Random/Grid)", "cost reduction via early stopping", "warm start reusing prior results". Today we cover search strategies, tuning job structure, and cost-reduction features—three axes.