Discussion of 'Space-filling designs for computer experiments: A review'

摘要

I commend Dr. Joseph for his excellent review of space-filling designs for computer experiments and his presentation of maximum projection designs from Joseph, Gul, and Ba (2015). His review article is both approachable to practitioners and thought-provoking for researchers. As research in the field of computer experiments matures, the need for practitioner-oriented guidance is essential, because computer experiments are not typically included in classical experimental design training. In addition, Dr. Joseph's review of space-filling designs complements the contributions of Zhang and Notz (2015) and Hig-don (2015) from the 2014 Stu Hunter Research Conference that examined modeling strategies for computer experiments. Computer experiments are ubiquitous across engineering and science disciplines, and they provide a powerful experimental strategy combined with physical experimentation. Computer experiments can reduce the amount of physical experimentation; enable faster, more flexible iteration between design and analysis; and provide the ability to explore a larger design space. Furthermore, their utility is continuously growing with increases in computational performance and higher fidelity models. In my discussion, I seek to bring a practitioner perspective by emphasizing several concepts from Dr. Joseph's review and proposing some challenges to motivate future research.