4.2 Genetic Algorithms

4.2 Genetic Algorithms continued...

GAs differ from traditional optimization algorithms in four important respects:

They work using an encoding of the control variables, rather than the variables themselves.
They search from one population of solutions to another, rather than from individual to individual.
They use only objective function information, not derivatives.
They use probabilistic, not deterministic, transition rules.

Of course, GAs share the last two attributes with SA and, not surprisingly, have found applications in many of the same areas.

The basic structure of a GA is shown in Figure 20. One minor change from the standard optimization routine flow diagram is the use of the word `population' rather than `solution.' A more major difference is that the usual operation of generating a new solution has been replaced by three separate activities --- population selection, recombination and mutation.

Figure 20: The Basic Structure of a Genetic Algorithm.

4.2.1 Solution Representation
4.2.2 Population Selection
4.2.3 Population Recombination
4.2.4 Population Mutation
4.2.5 Advanced Operators
4.2.6 Population Assessment
4.2.7 Control Parameters
4.2.8 GA Computational Considerations
4.2.9 GA Algorithm Performance