A methodology to develop Service Oriented Evolutionary Algorithms
García-Sánchez, Mora, Castillo, González and Merelo
Evolving solutions through natural selection
Select the best, combine and change them, repeat until solution found
From The descent of man by Darwin
Do current EA models have a problem?
Not integrated
Service oriented architectures offer
Language-independence
And distribution transparency
Standard service interfaces
Distributed control of nodes
SOMA methodology for service design
Identification
Problem, algorithm and infrastructure domains
Specification
Fitness, population, parameters, operators as services
Implementation and deployment
Choose technology for publication, discovery and distribution: OSGi/WSDL/other
Dynamic binding of services
Operators changed or added during runtime
EAs might fall in local minimum
Some selection operators have higher selection pressure
Image from Library of congress
Binary tournament: greedy
Roulette selection: relaxed
Identification and specification
Dynamic binding rules!
Services afford new ways of assembling evolutionary algorithms
Dynamic monitoring and binding of operators/services rock!