Advantages and risks of genetic algorithms
It should be clear by now that genetic algorithms provide scientists with a powerful optimization tool for problems that:
Are poorly understood.
May have more than one good enough solution.
Have discrete, discontinuous, and non-differentiable functions.
Can be easily integrated with the rules or policies engine (see the Learning classifiers systems section in Chapter 15, Reinforcement Learning).
Do not require deep domain knowledge. The genetic algorithm generates new solution candidates through genetic operators without the need to specify constraints and initial conditions.
Do not require knowledge of numerical methods such as the Newton-Raphson, conjugate gradient, or L-BFGS as optimization techniques, which frighten those with little inclination for mathematics.
However, evolutionary computation is not suitable for problems for which:
A fitness or scoring function cannot be quantified or even defined
There is a need to find the global minimum or maximum...
