CODEWITHSUNDEEP
genetic-algorithmgenetic-programmingparticle-swarm
Arrofi Reza Satria
Arrofi Reza Satria

Reputation: 190

How many offspring should we produce in Genetic Algorithm?

Suppose I have 100 popsize, should I make 10 offspring? I want the best combination between popsize and offspring to achieve convergent quickly and please also include the paper.

Upvotes: 2

Views: 1578

Answers (2)

doplano
doplano

Reputation: 1581

-Suppose N population size for your GA.

chrom # 0 = "01010110101" | Fitness = f0
chrom # 1 = "11010010111" | Fitness = f1
chrom # 2 = "01010111011" | Fitness = f2
chrom # 3 = "01111010100" | Fitness = f3
.
.
.
chrom # N = "01011010110" | Fitness = fN

-You apply tournament of chromosome randomly from the main population with size T : (T < N)

Tournament chrom # 0 = "01010110101" | Fitness = f0
Tournament chrom # 1 = "11010010111" | Fitness = f1
Tournament chrom # 2 = "01010111011" | Fitness = f2
Tournament chrom # 3 = "01111010100" | Fitness = f3
.
.
.
Tournament chrom # T = "01011010110" | Fitness = fT

to simply get mate chromosome:

Mate Chromosome # 1

another Tournament :

Mate Chromosome # 2

you apply crossover to return offspring:

Crossover(Mate Chromosome # 1, Mate Chromosome # 2) => offspring

That technically means you'd get N offspring for your new population.

Mutation(offspring) => new chromosome for new population

Continue the iteration until you converge to the max size of the target chromosome.

Upvotes: 2

J63
J63

Reputation: 843

There is not best offspring/population to every problem you can solve with genetic algorithm. Every problem has a best configuration, which is unknown , not only of offspring/population but also of mutation chance, chromosome design, etc...

With that said there are multiple solutions like total substitution, partial substitution, etc.. With it's own benefits and downsides. You should explore them and decide which one fits your problem the most.

Upvotes: 2

Related Questions