Parallel genetic algorithms for large-scale fixed charge networks
We present parallel genetic algorithms (GA`s) for several classes of fixed-charge multicommodity flow problems arising from applications in parallel database design, domain decomposition, and telecommunications. These algorithms utilize a high-level approach based upon representing individual (in the GA sense) in terms of selections from a library of pre-computed {open_quotes}building blocks{close_quotes} of sets of variables rather than as values of individual binary variables corresponding to single links. The fitness function for this form of representation is then evaluated by applying heuristics to the starting point represented by an individual, thereby allowing for modifications in the original {open_quotes}blueprint{close_quotes} represented by the individual. These heuristics lead to objective function improvements and are also used to force feasibility. With this type of fitness function, the amount of time spent on the other operations of the GA (selection, mutation, etc.) is relatively small, so that high efficiency may be achieved in parallel implementations of the algorithm. We present computational results on the CM-5 supercomputer, demonstrating the ability to solve to optimality certain fixed-charge problems with more than one million binary variables.
- OSTI ID:
- 36630
- Report Number(s):
- CONF-9408161-; TRN: 94:009753-0609
- Resource Relation:
- Conference: 15. international symposium on mathematical programming, Ann Arbor, MI (United States), 15-19 Aug 1994; Other Information: PBD: 1994; Related Information: Is Part Of Mathematical programming: State of the art 1994; Birge, J.R.; Murty, K.G. [eds.]; PB: 312 p.
- Country of Publication:
- United States
- Language:
- English
Similar Records
A parallel genetic algorithm for the set partitioning problem
Entropy based primal-dual algorithm for convex and linear cost transportation problems with serial and parallel implementations