Assess students' ability to apply Bayesian optimization with Gaussian-process surrogates and acquisition functions (e.g., EI, PI, UCB) to tune deep neural network hyperparameters under computational/resource constraints. Scope: kernel choice, GP scalability (sparse/approximate), batch/parallel and cost-aware or multi-fidelity acquisition strategies, early-stopping, and evaluation by sample efficiency and wall-clock/time-to-best.