What is the significance of the mutation rate in genetic algorithms?
Hello. Let's say I set a mutation rate of 5% in an AG. What does that mean?
- What do I iterate over the entire population and, for each individual, iterate over all of his genes and make a change with 5% probability? Here, the entire population will have about 5% of their genes changed.
- What do I iterate over the entire population and, for each individual, do I change a random gene from it with 5% probability? Here, at the end of the mutation, about 5% of the population will have an altered gene.
Something else...Conventionally, is the mutation applied to the entire population, including the selected parents, or only to the offspring? The second option is biologically more plausible.
1 answers
The mutation rate refers to the amount of individuals in the population that will undergo mutation.
It is usually very small and serves to maintain the genetic variability of chromosomes as generations pass.
You can, of course, arbitrate another parameter in your implementation that controls the percentage of genes that will be mutated during the process.
In relation to the other question, in general, the mutation comes right after crossing. Thus, it acts only on offspring.
I suggest the Manual of Evolutionary Computation and metaheuristics as an introduction to the topic.