| edurant.com > Papers > Genetic Algorithm Introduction |
Genetic algorithms (GAs) are a class of algorithms for maximizing an objective function. They exploit the (not necessarily continuous) structure of the error surface. GAs do not assume that the error surface is unimodal, or even that its derivative exists. Such assumptions are required for efficient use of traditional optimization strategies. However, many practical design problems are nonlinear and multimodal, making the GA attractive.
A genetic algorithm performs a parallelized stochastic search. It is parallelized in that many solutions ("genes") are considered simultaneously (constituting a "population"). It is stochastic in that (i) solutions are chosen for refinement randomly, with the probability of selection enhanced by the quality of the solution ("fitness") and (ii) the search direction about a solution (or between solutions) is chosen randomly.