1) What is the minimum number of benchmark functions we are supposed to test on our newly develop crossover and mutation operators? If we want to compare the performance of 2 operators, must we use the same benchmark functions to test? How to identify the appropriate benchmark functions to test our operators? Or do I randomly test it on some standard benchmark functions? How to know which benchmark function is to test which category of operators?
2) There are many crossover and mutators introduced, what are the main features of these operators we must know to ensure that comparison among the operators are compatible? If I have a new discrete mutator, can I compare it with a polynomial mutator? If I have a new PCA mutator, can I compare it with a real coded mutator?
3) Is there a particular crossover and mutator to solve a particular optimization problem? What are the characteristics of the operators and the characteristics of the fitness functions which I must know to ensure that they are compatible to apply?
4) What are the main differences between the benchmark functions of scalable, non-scalable, unimodal and multimodal?