What is it about?
Engineering optimization problems are extremely challenging due to their strongly nonlinear nature and expensive evaluation function, making them hard for optimization algorithms. This work proposes an approach towards learning the landscape characteristics of engineering optimization, using Exploratory Landscape Analysis (ELA), with special emphasis on automotive crash. With our approach, objective functions with characteristics similar to real-world engineering design problems, but are fast-to-evaluate, can be systematically generated.
Featured Image
Photo by Kalen Emsley on Unsplash
Why is it important?
Having an additional insights into the landscape characteristics of an optimization problem could potentially allow us to have a better understanding on optimization algorithm performances. In other words, by knowing the optimization landscape characteristics, we can tailor the best suited algorithm for our problems. To go one step further, objective functions with similar characteristics can be exploited as "proxy-function" in tuning optimization algorithms for real-world problems.
Perspectives
This work opens up a new approach in handling expensive real-world optimization problems. I am truly fascinated by the fact, that our approach is, theoretically and ideally, applicable and extensible to all kind of engineering domains, and not just in automotive crash. We also have the honor to receive Best Paper Award in The Genetic and Evolutionary Computation Conference (GECCO) 2022.
Fu Xing Long
Universiteit Leiden
Read the Original
This page is a summary of: Learning the characteristics of engineering optimization problems with applications in automotive crash, July 2022, ACM (Association for Computing Machinery),
DOI: 10.1145/3512290.3528712.
You can read the full text:
Resources
Contributors
The following have contributed to this page
