Hands-On Genetic Algorithms with Python

Serial niching versus parallel niching

Unfortunately, the niching concept as described previously can prove hard to implement as it increases the complexity of the fitness calculation. In practice, it will also require the population size to be the original one multiplied by the number of the expected peaks (which is generally unknown).

One way to overcome these issues is to find the peaks one at a time (serial niching) instead of attempting to find all of them at the same time (parallel niching). To implement serial niching, we use the genetic algorithm as usual and find the best solution. We then update the fitness function so that the area of the maximum point that was found is flattened, and repeat the process of the genetic algorithm. ...