Genetic algorithm, confusion with selection/reproduction step.

[Hans Rudel]

Hi, im a little confused with the selection/reproduction step in genetic algorithms. Im aware that the initial population has several hundred in it. That the fitness function is used to select a certain number of the possible solutions for cross over/mutation. What i dont get is how the population size, between each successive generation, doesnt get progressively smaller and smaller?

Thanks for your time.

 

[Dason]

That's what the reproduction step is for. So after you choose the best children in one step you use those as parents to create an appropriate number of children to get back to a population of appropriate size.

http://en.wikipedia.org/wiki/Genetic_algorithm#Reproduction

 

[Hans Rudel]

Hi Dason, thanks very much for replying. So say i had 300 possible solutions in the initial population, and i selected 50 with the best fitness function result, Do i just keep selecting, based on their fitness result, pairs (or X number of parents as wiki suggests) until i have a new population of 300?

Are there any cases where you would want a large population to begin with but then narrow it down, ie 300 down to 100.

Thanks again for your help, i really appreciate it.

 

[Dason]

I won't lie - I really only know about genetic algorithms theoretically. I've never actually implemented one so I'm probably not the best to ask about implementation details.

 

[Hans Rudel]

no worries, i'll assume i use those 50 with the highest fitness to make a new population of 300 unless someone else can say otherwise. Thanks again for ur ehlp though, i really appreciate it.