Design of a parallel genetic algorithm for the Internet

This paper proposes a parallel implementation of the genetic algorithm (GA) on the Internet which will improve the algorithm´s performance. It is motivated by the possibility of aiding research into complex search and optimization problems that use the GA. Requirements and constraints regarding parallelization of the GA are identified. A parallel GA is developed for an ideal PRAM architecture and is shown to have an asymptotic running time of O(log n), an improvement over the sequential GA. A parallel GA is also designed for a Unix network and has an asymptotic running time comparable to the ideal system. The algorithm is a decentralized, asynchronous, and fault-tolerant design that matches the characteristics of the network. The GA population is divided into colonies that are distributed among processors. Trade policies are executed for the exchange of genes