Optimal Experimental Design

Optimal experimental design (OED) focuses on strategically selecting experiments to maximize information gain about a system or model, minimizing cost and effort. Current research emphasizes computationally efficient methods, particularly leveraging deep learning architectures like neural networks and normalizing flows, and Bayesian approaches to handle uncertainty and incorporate prior knowledge, often targeting specific objectives or quantities of interest. These advancements are improving the efficiency and robustness of experiments across diverse fields, from materials science and engineering to clinical trials and online platform A/B testing, leading to more reliable and insightful results.