# Genetic Algorithms: Choosing a Fitness Function

1. Apr 7, 2012

### ashmai

Does anyone know of a good fitness function for evaluating an animal characteristic in genetic algorithms?
I'm doing a project. The prompt is: Select some features of an actual organism and design a genetic algorithm together with a reasonable fitness landscape to evolve the organism in time.
I'm using the Matlab genetic algorithms toolbox, but can't think of a good organism characteristic that has an interesting fitness function. I can do anything, like bird beak length, elephant ear radius, etc... I just need a good way to evaluate how the possible genotypes will survive - thus, I need a good fitness function! Maybe from literature?

2. Apr 7, 2012

### Office_Shredder

Staff Emeritus
There are two possibilities:

1) The characteristic you picked is always better to have more of. FOr example, you might argue the longer the bird beak the better, and that's that. Then a fitness function like f(length)=length or f(length) = 1-1/(length) would work

2) It can be bad for it to be too big. For example if the beak's length is too long maybe the bird can't fly around anymore. So the fitness function is a quadratic function fitness = length*(10-length).

Obviously I can't be sure without knowing what the rest of your project says but when they say 'reasonable function' it sounds more like they just want you to pick a function that sounds pretty good rather than actually calculating one from empirical data