What is it about?
This paper proposes a new optimization problem for processes that require expensive physical experiments to produce candidate solutions. The solutions require resources that can be bought, produced, and stored, and are shared across the solution space. This new optimization problem searches for a close-to-optimal solution while minimizing the overall use of resources throughout the production process. We test this new optimization problem in the field of molecular discovery.
Featured Image
Photo by Girl with red hat on Unsplash
Why is it important?
Resource-dependent evaluations make exploitation cheaper and exploration more expensive than a typical optimization problem. The production and evaluation of physical candidate solutions often have a surplus of resources that can be reused later. Neighboring solutions that require resources already in storage are cheaper to evaluate. Thus, solutions are no longer independent in evaluation when resources are costly and a good solution must be found under a budget. Our paper proposes different optimization algorithms to address this new optimization problem.
Perspectives
This article is an important first step towards a new problem class in the popular field of drug discovery. I believe the methodology that we have used to model the production process and problem as a whole can be used in many other fields where candidate solutions have to be produced via lengthy and expensive processes.
Stefan Pricopie
University of Manchester
Read the Original
This page is a summary of: Expensive optimization with production-graph resource constraints, July 2022, ACM (Association for Computing Machinery),
DOI: 10.1145/3512290.3528741.
You can read the full text:
Contributors
The following have contributed to this page
