1

I'm wondering if I can select two same parents in two iteration of selection in a genetic algorithm (in a same population with tournament selection).

Can I?

manlio
  • 18,345
  • 14
  • 76
  • 126
Farhaneh Moradi
  • 109
  • 1
  • 1
  • 10

1 Answers1

1

A lot of these decisions are made after experimentation with one's particular software and domain.

Of course two parents can generate more than two children. This may happen either because:

  • crossover operator creates more than two children;
  • tournament selection picks repeatedly the same parents (with a simple steady state population this is a common event).

Generally it's not recommended to create too many individuals with the same parents because you could have a too "restricted trend" (what "too many" means is debatable).

So you can often find some form of prevention. Apart from checking explicitly for the "same parents" occurrence, there are other techniques.

E.g.

  • demetic grouping the same parents can generate a numerous offspring but children will compete among them.

  • family competition replacement schemes are different way of limiting the amounts of multiple crossovers with the same parents.

  • ...

manlio
  • 18,345
  • 14
  • 76
  • 126
  • Thank you and one more question: what are the reasons of getting nearly same population in each generation? does one point cross over cause this problem? I used tournament selection and one point cross over with probability 0.8 and mutation by probability 0.02. – Farhaneh Moradi May 30 '15 at 18:58
  • @FarhanehMoradi It's hard to say without more data. It seems a [premature convergence](http://en.wikipedia.org/wiki/Premature_convergence) issue. You could try using uniform crossover, a higher mutation probability, a bigger population, some form of fitness sharing or the algorithms named in the answer. – manlio May 30 '15 at 20:31