A new paper proposes restructuring Bayesian optimization through linear structure, making it up to three orders of magnitude faster while maintaining optimization performance comparable to state-of-the-art methods.