|
|
In this work we invesigate a new and broad class of constrained
optimization problem. An ephemeral resource constraint (ERC) is a
temporary limitation on the availability of resources needed
to evaluate some subset of the solution candidates (a subspace) in an optimization
problem. ERCs arise when doing closed-loop optimization because
equipment, personnel, or consumables needed to make or test solutions
may be in limited supply, or be subject to failure or unavailability.
In this work, we formulated ERC problems for the first time, created
benchmarks,
and investigated the effect of ERCs on evolutionary optimization. We then
designed several strategies to mitigate slowing of evolutionary
progress.
Our investigations of this broad problem class are just at the beginning.
Papers
- R. Allmendinger and J. Knowles (2013) On Handling Ephemeral
Resource Constraints in Evolutionary Search. Evolutionary
Computation, 21(3): 497-531. Download PDF
(Pre-publication); Published version at
MIT
Press
- Allmendinger, R. and Knowles, J. (2010) On-Line Purchasing
Strategies for an Evolutionary Algorithm Performing
Resource-Constrained Optimization Proceedings of PPSN
XI, volume II, LNCS 6239, pp
161-170. GS
link
[ Back ] to Research Main Page
|