21

In most implementations of genetic algorithms, the focus is on crossover and mutation. But somehow, most of them leave out diploid (dominant/recessive) nature of genes. As far as my (limited) understanding goes dominant/recessive nature of genes is a very important factor in deciding the actual characteristics of an organism.

So my question is why is the diploid nature of genes left out of genetic algorithms in most implementation?

Is it because:

  • it doesn't provide much benefit
  • it adds unnecessary complexity to an otherwise simple algorithm
  • it's difficult to implement

Or something else entirely?

Shayan RC
  • 633
  • 1
  • 5
  • 11

2 Answers2

17

I don't know the actual reason, but it feels intuitive: let's think about what the diploid nature of genes does in RL. In essence it allows the recessive gene to remain in the gene pool even if it's currently at disadvantage to exist, and occasionally resurface, giving two things - first, it doesn't go extinct and can re-multiply if it becomes advantageous; and second, it ensures some population variety as you'll continuously have both phenotypes - a part of population that exhibits the gene and part that doesn't.

Both of these things can be achieved in simpler ways by the mutating/crossover engine - you can directly 'fetch' random good-performing items from 100000 generations ago (which nature usually can't); and you can keep multiple varied subpopulations while protecting the non-primary ones from extinction, which nature usualy doesn't do.

Peteris
  • 426
  • 4
  • 9
1

See Occam's Razor

Among competing hypotheses, the one with the fewest assumptions should be selected. Also: Entities must not be multiplied beyond necessity.

If both hypotheses are equally good, choose the more simple version because the more complex version makes assumptions about something which you cannot be sure.

The question is, do diploid dominant and recessive genes provide more functionality that allows us to describe a richer hypothesis space?

  • Could we do something that cannot be achieved through simple mutation? No. Mutation can create any new sequence.
  • Could we do something with mutation that cannot be achieved through diploid dominant and recessive genes? Yes. Mutation allows any random new sequence while diploid genes would only recover something seen before and lost.

The only possible benefit left to explore is whether or not diploid genes would somehow be more efficient. It seems from their lack of use that this isn't the case. Mutations are usually small changes in an answer. The benefit of keeping a past, good answer around is small. It can easily pop up again.

Biology can be used as inspiration for computer models, but it rarely has the best answer. Biology generates solutions by chance and natural selection where DNA is concerned. Biology is also solving different problems with different raw materials and tools. Look at the way birds and bats fly. Why aren't our airplanes designed to move their wings up and down to take off or move higher? Because it would be horribly inefficient. Jet propulsion and helicopters are more suited to our needs. We can carry heavier payloads and travel at much faster speeds than birds and bats.

Charity Leschinski
  • 111
  • 3