In particular, in the fields of genetic programming and genetic algorithms, each design solution is commonly represented as a string of numbers. After each round of testing, or simulation, the idea is to delete the n worst design solutions, and to breedn new ones from the best design solutions. Each design solution, therefore, needs to be awarded a figure of merit, to indicate how close it came to meeting the overall specification, and this is generated by applying the fitness function to the test, or simulation, results obtained from that solution. The reason that genetic algorithms cannot be considered to be a lazy way of performing design work is precisely because of the effort involved in designing a workable fitness function. Even though it is no longer the human designer, but the computer which comes up with the final design, it is still the human designer who has to design the fitness function. If this is designed badly, the algorithm will either converge on an inappropriate solution, or will have difficulty converging at all. The fitness function should not only correlate closely with the designer's goal, but it also should be computationally efficient. Speed of execution is very important, as a typical genetic algorithm must be iterated many times in order to produce a usable result for a non-trivial problem. Fitness approximation may be appropriate, especially in the following cases:
Two main classes of fitness functions exist: one where the fitness function does not change, as in optimizing a fixed function or testing with a fixed set of test cases; and one where the fitness function is mutable, as in niche differentiation or co-evolving the set of test cases. Another way of looking at fitness functions is in terms of a fitness landscape, which shows the fitness for each possible chromosome. Definition of the fitness function is not straightforward in many cases and often is performed iteratively if the fittest solutions produced by genetic algorithms are not what is desired. Interactive genetic algorithms address this difficulty by outsourcing evaluation to external agents which are normally humans.
