Genetic Algorithms: Fitness Function and Selection
The fitness function can be defined as a particular solution to a particular problem through corresponding input and produces output as to how good the solution is with respect to the given problem.
The fitness value calculation is done repeatedly in a genetic algorithm and that is why it must be fast enough. Slow problem-solving computation can have an effect on the genetic algorithm and can make it too slow to execute.
In some cases, the genetic fitness function and function objects are nearly equal or the same because the function objective is to either minimize or maximize some complex computation problems with multiple constraints. A designer algorithm must be used to have a different fitness function.
The following characteristics must be produced by the fitness function :
- It must be fast to easily solve or compute complex problems.
- It must measure in quantity how good or fit the function provided or how much good can it provide individuals from the particular solution.
Some cases may happen that a fitness function cannot calculate the given complex problem directly thereafter approximation of fitness function may be used to get with the correct solution, and that problem may occur because of inherit in complexities of the problem.
USES OF GENETIC FITNESS FUNCTIONS:
- We test with complex problems and come up with the best case solution.
- Each solution is given a particular approximation to check whether it is near the original value or not.
REQUIREMENTS OF A FITNESS FUNCTION :
- It must be clearly defined so that it can be understood how efficiently the fitness is being calculated.
- It must be efficiently produced executed or implemented. If the function becomes the congestion of the algorithm then the efficiency of the overall algorithm may be lowered.
- The particular fitness function must produce an approximated solution.
TO OVERCOME A PARTICULAR FITNESS FUNCTION FOR A GIVEN PROBLEM :
Every particular problem has its own fitness function. There is no particular rule to use a fitness function. However, a few methods are being adopted by scientists that deal with data.
For optimization of a particular complex problem, functions that are basic must be used such as the total set of calculations of parameters that are directly associated with the complexity of the problem and its domain can be used to solve in a fitness function.
A fitness function must be such a kind of function that can measure how fast or better it can produce solutions for a given set of problems. For example, suppose a fitness function is almost zero unless a function produces it is correct or not is not better, since it does not provide the data how approximate is the solution to the correct answer.
Similarly, a fitness function that is increasing in nature will give solutions better but it would not identify or reallocate the solution which is best is not approximately close enough to the actual answer.
Thus, the fitness function is also known as the evaluation function which evaluates how close or approximate it is or how nearly it is to the original solution.
Reference