Publication detail

The Nested Genetic Agorithms for Distributed Optimization Problems

ROUPEC, J. POPELA, P.

Original Title

The Nested Genetic Agorithms for Distributed Optimization Problems

English Title

The Nested Genetic Agorithms for Distributed Optimization Problems

Type

conference paper

Language

en

Original Abstract

Firstly, we review basic principles of the distributed modeling approach in optimization and present introduction to the formal framework based on the concept of a distributed optimization program. The framework is a general one and may be utilized for various classes of decision problems. The DOPs (distributed optimization programs) are introduced as syntactical entities containing certain optimization elements and based on composition rules. They may describe both basic and advanced mathematical programs (e.g., dynamic, stochastic, multistage, and hierarchical) and also game theory models. In addition, more complicated models can be derived from these building stones and further transformed in the syntactical correct way. Although the introduced descriptions are particularly designed for manipulations of programs structures, semantics for certain DOPs can also be defined. Hence, the next challenge is to search promising solutions in the feasible sets of optimization elements of DOPs. Therefore, several genetic algorithms (GAs) are chosen to search in separate feasible sets and they may also exchange information about different populations for achieved solutions of DOP elements in various ways. The general inspiration comes from decomposition techniques in scenario-based multistage programs, so the name nested GAs is used in our case. The computational results and implementation description are presented for the specific min-max problems that are chosen as elementary prototype instances.

English abstract

Firstly, we review basic principles of the distributed modeling approach in optimization and present introduction to the formal framework based on the concept of a distributed optimization program. The framework is a general one and may be utilized for various classes of decision problems. The DOPs (distributed optimization programs) are introduced as syntactical entities containing certain optimization elements and based on composition rules. They may describe both basic and advanced mathematical programs (e.g., dynamic, stochastic, multistage, and hierarchical) and also game theory models. In addition, more complicated models can be derived from these building stones and further transformed in the syntactical correct way. Although the introduced descriptions are particularly designed for manipulations of programs structures, semantics for certain DOPs can also be defined. Hence, the next challenge is to search promising solutions in the feasible sets of optimization elements of DOPs. Therefore, several genetic algorithms (GAs) are chosen to search in separate feasible sets and they may also exchange information about different populations for achieved solutions of DOP elements in various ways. The general inspiration comes from decomposition techniques in scenario-based multistage programs, so the name nested GAs is used in our case. The computational results and implementation description are presented for the specific min-max problems that are chosen as elementary prototype instances.

Keywords

Genetic algorithms, minmax problems, distributed optimization programs, nested decomposition

RIV year

2011

Released

19.10.2011

ISBN

978-988-18210-9-6

Book

Proceedings of The World Congress on Engineering and Computer Science 2011

Pages from

480

Pages to

484

Pages count

5

BibTex


@inproceedings{BUT75181,
  author="Jan {Roupec} and Pavel {Popela}",
  title="The Nested Genetic Agorithms for Distributed Optimization Problems",
  annote="Firstly, we review basic principles of the distributed modeling approach in optimization and present introduction to the formal framework based on the concept of a distributed optimization program. The framework is a general one and may be utilized for various classes of decision problems. The DOPs (distributed optimization programs) are introduced as syntactical entities containing certain optimization elements and based on composition rules. They may describe both basic and advanced mathematical programs (e.g., dynamic, stochastic, multistage, and hierarchical) and also game theory models. In addition, more complicated models can be derived from these building stones and further transformed in the syntactical correct way. Although the introduced descriptions are particularly designed for manipulations of programs structures, semantics for certain DOPs can also be defined. Hence, the next challenge is to search promising solutions in the feasible sets of optimization elements of DOPs. Therefore, several genetic algorithms (GAs) are chosen to search in separate feasible sets and they may also exchange information about different populations for achieved solutions of DOP elements in various ways. The general inspiration comes from decomposition techniques in scenario-based multistage programs, so the name nested GAs is used in our case. The computational results and implementation description are presented for the specific min-max problems that are chosen as elementary prototype instances.",
  booktitle="Proceedings of The World Congress on Engineering and Computer Science 2011",
  chapter="75181",
  howpublished="print",
  year="2011",
  month="october",
  pages="480--484",
  type="conference paper"
}