tournamentSelection {nsga2R} | R Documentation |
Tournaments are played among several solutions. The best one is chosen according to their nondomination levels and crowding distances. And it is placed in the mating pool.
tournamentSelection(pop, pool_size, tour_size)
pop |
Population matrix with nondomination rank and crowding distance |
pool_size |
Size of mating pool, usually same as the population size |
tour_size |
Size of tournament, the selection pressure can be adjusted by varying the tournament size |
Return the mating pool with decision variables, objective functions, nondomination level, and crowding distance
Ching-Shih (Vince) Tsou cstsou@mail.ntcb.edu.tw
Deb, K., Pratap, A., Agarwal, S., and Meyarivan, T. (2002), " A fast and elitist multiobjective genetic algorithm: NSGA-II", IEEE Transactions on Evolutionary Computation, 6(2), 182-197.
library(mco)
tourSize <- popSize <- 10
lowerBounds <- rep(0,30)
upperBounds <- rep(1,30)
varNo <- length(lowerBounds)
objDim <- 2
set.seed(1234)
population <- t(sapply(1:popSize, function(u) array(runif(length(lowerBounds),
lowerBounds,upperBounds))))
population <- cbind(population, t(apply(population,1,zdt3)))
ranking <- fastNonDominatedSorting(population[,(varNo+1):(varNo+objDim)])
rnkIndex <- integer(popSize)
i <- 1
while (i <= length(ranking)) {
rnkIndex[ranking[[i]]] <- i
i <- i + 1
}
population <- cbind(population,rnkIndex);
objRange <- apply(population[,(varNo+1):(varNo+objDim)], 2, max) -
apply(population[,(varNo+1):(varNo+objDim)], 2, min);
cd <- crowdingDist4frnt(population,ranking,objRange)
population <- cbind(population,apply(cd,1,sum))
matingPool <- tournamentSelection(population,popSize,tourSize)
matingPool