CODEWITHSUNDEEP
mathematical-optimizationgenetic-algorithmevolutionary-algorithmgenetic-programming
Akhilnandh Ramesh
Akhilnandh Ramesh

Reputation: 27

A query regarding NSGA 2: Tournament selection

In NSGA 2 algorithm, we employ binary tournament selection as the selection procedure to generate mating parents from a population. The offspring data set's size is the same as the size of the parents. I have a question regarding this. I am new to GAs and this may sound absurd to a few more experienced users out there, but any help will be much appreciated.

Initially, a parent population Po is created and sorted on the basis of nondomination. Binary tournament, crossover, and mutation are utilized to create the offspring population of the same size as the parent.

The binary tournament selection is used for selecting the mating pool.

The offspring population is of the same size as the parent, which means mating pool is of the same size as the parent set and would contain all elements of the parent's set.

Then, in this case, why would we need to adopt a tournament selection? we can sort the parent set on the basis of descending fitness values and just ensure crossover and mutation are carried out with populations in decreasing order of fitness.

Should the initial population be greater than the parent population size?

The original paper by Dr K.Deb- on NSGA 2: https://www.iitk.ac.in/kangal/Deb_NSGA-II.pdf

Does not mention that we need a population set greater than the parent population set for the mating to take place

Upvotes: 1

Views: 535

Answers (1)

33fred33
33fred33

Reputation: 67

The only thing you got wrong is that the individuals are replaced.

A single individual can be selected multiple times as a parent (tournament selection is made with replacement).

The best individual is expected to be selected many times, a phenomenon called "selection pressure". This is desired to spread the best individual's genes to the offspring population.

For the sake of completeness: You don't always want the crossover to happen between similar individuals, it could lead to a lack of diversity. The sorting idea wouldn't be ideal.

Upvotes: 1

Related Questions