dominance change mechanism, performance improvedsignificantly. Further, extending the additive dominancescheme with change improved the performanceconsiderably . They concluded that some form of dominance mechanism is needed along with diploidy toallow flexible response to change.In 2002, S. Ayse, Yilmaz, Annie S. Wu, proposed a newdiploid scheme without dominance on integerrepresentation. In their research, they evolved all diploidindividuals without using any haploid stage and comparedperformance for TSP as their test problem. They concludedthat simple haploid GA outperformed diploid GA . In2003, C. Ryan, J.J. Collins, D. Wallin extended their work and proposed the Shades scheme – a new version of haploidy that incorporates the characteristics of diploidscheme in haploid genetic algorithms but with lesser cost.Performance of Shades scheme was analyzed andcompared to two diploid schemes – tri-allelic andDominance change mechanism scheme in two dynamicproblems domains. Shades-3 outperformed both the diploidschemes in both Osmera’s dynamic problem andconstrained knapsack problem. Shades-2 outperformedshades -3 in knapsack problem .In 2003, Robert Schafer presented a GA protocol as atool to approach dynamic systems having reciprocalindividual-environment interaction and then applied on amodel problem in which a population of simulatedcreatures lived and metabolized in a three-gas atmosphere. In 2005, Shane Lee and Hefin Rowlands described adiploid genetic algorithm, which favoured robust localoptima rather than a less robust global optimum in aproblem space. Diploid chromosomes were created withtwo binary haploid chromosomes, which were then used tocreate a schema. The schema was then used to measure thefitness of a family of solutions. .In 2007, Shengxiang Yang proposed an adaptivedominance learning scheme for diploid genetic algorithmsin dynamic environments. In this scheme, the genotype tophenotype mapping in each gene locus was controlled by adominance probability . The proposed dominancescheme was experimentally compared to two other schemesfor diploid genetic algorithms and results validated theefficiency of the dominance learning scheme. Out of thetwo schemes, additive diploidy scheme proved to be betterthan the Ng-Wong dominance scheme. In 2009, Dan Simonutilized diploidy and dominance in genetic algorithms toimprove performance in time-varying optimizationproblems. He used the scaled One Max problem to provideadditional theoretical basis for the superior time-varyingperformance of diploid GAs. The analysis confirmed thatdiploidy increases diversity, and provided somequantitative results for diversity increase as a function of the GA population characteristics .IV.
The Travelling Salesman Problem (TSP) is a classicalcombinatorial optimization problem, which is known to beNP-Hard problem. The problem is to find the shortest touror Hamiltonian path through a set of N vertices so that eachvertex is visited exactly once . To find an optimalsolution involves searching in a solution space that growsexponentially with number of cities. So, certain kind of heuristic techniques are applied to reduce the search spaceand direct search to the areas with the highest probability of good solutions. One such heuristics technique is geneticalgorithms (GAs).The problem is solved under following assumptions:
Each city is connected to every other city,
Each city has to be visited exactly once,
The salesman’s tour starts and ends at the samecity.Based on the above assumptions, a simple geneticalgorithm is formulated to solve the problem.
[N is number of cities and M is number of maximumgenerations, GP is generation pool ]1
Create an initial population P(i) of GPchromosomes having length N.4
Evaluate the fitness of each chromosome in P(i).5
While i <M do6
Perform selection i.e. choose at random apair of parents from P(i).7
Exchange strings by crossover to createtwo offsprings.8
Insert offsprings in P(i+1)9
Repeat steps 6 to 8 until P(i+1) is full10
Replace P(i) with P(i+1).11
Evaluate the fitness of each chromosomein P(i+1)12
Final result is best chromosome created duringthe search.14
The algorithm is further coded in MATLAB for itsimplementation using both haploid and diploid genome set.The code was implemented first for 10 cities. The cost of different paths was computed for fifteen consecutive runsand then compared.
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No. 7, October 2010236http://sites.google.com/site/ijcsis/ISSN 1947-5500