Open rochaporto opened 9 years ago
This is a really good description: http://natureofcode.com/book/chapter-9-the-evolution-of-code/
And as a summary...
Population. Start with a variable number of random candidates
Fitness. Function evaluating the gene's fitness. Distance, prob squared or 2^distance so that we get an exponential increase for better solutions
Selection. Based on fitness. Should be a roulette wheel with the fittest candidates having a higher selection probability
Reproduction. Crossover between 1, 2 or even more parents. Take random position and take each side from a different parent.
Mutation. For each point mutate based on a probability, which should be low. Guarantees randomness when start population was bad
Play with population number and mutation rate to evaluate best results.
Genotype is the internal representation of a candidate. Phenotype the visual one
Another good pointer. http://www.obitko.com/tutorials/genetic-algorithms/ga-basic-description.php
This is a really good description: http://natureofcode.com/book/chapter-9-the-evolution-of-code/
And as a summary...
Population. Start with a variable number of random candidates
Fitness. Function evaluating the gene's fitness. Distance, prob squared or 2^distance so that we get an exponential increase for better solutions
Selection. Based on fitness. Should be a roulette wheel with the fittest candidates having a higher selection probability
Reproduction. Crossover between 1, 2 or even more parents. Take random position and take each side from a different parent.
Mutation. For each point mutate based on a probability, which should be low. Guarantees randomness when start population was bad
Play with population number and mutation rate to evaluate best results.
Genotype is the internal representation of a candidate. Phenotype the visual one