One option for applying mutation in real-coded genetic algorithms is to replace any real value with a brand new one, generated randomly. However, this can result in a mutated individual that has no relationship to the original individual.
Another approach is to generate a random real number that resides in the vicinity of the original individual. An example of such a method is the normally distributed (or Gaussian) mutation: a random number is generated using a normal distribution with a mean value of zero and some predetermined standard deviation as shown in the following plot:
Gaussian mutation distribution example
In the next ...
