I have been searching for genetic algorithm a lot, But I could not understand necessity of using crossover and mutation simultaneously yet. In an online course, following paragraph was written:
crossover is an operation which drive the population towards a local maximum(or minimum). If we use only crossover, it will yield approximately the same result as hill-climbing algorithm!!!mutation is a so-called divergence operation force one or more individuals of the population to discover other regions of the search space. So, this is essential in order to find the global optimum.
I can not understand it easily, especially because in metaheuristic algorithms, we must cope with somehow statistical-based optimization. Moreover, I had implemented GA in python and still can not realize the performance difference between these two parameters.
You can have a look at this paper for your interest
Article Metaheuristic Algorithms in Optimal Power Flow Analysis: A Q...
There are two concepts related to optimization.
1) Exploration
2) Exploitation
Through crossover you are exploiting (mostly) or searching solution within the found solutions i.e. new chromosomes are formed using same values (genes) of the chromosomes. Whereas mutation can help to bring some more diversity in the population and can mutate a gene to any random point. Hence, mutation can introduce more diversity through exploration of other solutions rather than existing solutions.
Mohammad Reza Moravej
- Badges
- Science topic