Fundamentals of Statistics contains material of various lectures and courses of H. Lohninger on statistics, data analysis and chemometrics......click here for more. |
Home Multivariate Data Optimization Survey of Methods Genetic Algorithms | |||||||||
See also: optimization methods | |||||||||
Optimization Methods - Genetic AlgorithmsVarious attempts have been made to combine the advantages of deterministic and random-search methods. One particular approach has been investigated in recent years: genetic algorithms. The idea behind these methods is to exploit the principles of genetics for the optimization theory. First of all, a population of "explorers" is created. These explorers are positioned at random within the search space (phase space). Each "explorer" (in genetic algorithm terminology: individuum) detects the value of the response function at its own location and feeds it to a fitness function. The fitness is designed in a way that it maximizes when the search goal is reached. Depending on the value of the fitness function, several basic operations are performed:
These steps are repeated, until some termination criterion is fulfilled (e.g. the best explorer reached some defined threshold of fitness), or a defined number of generations have been calculated. The second strategy is often applied if no information about the global optimum is available. The most important advantage of genetic algorithms is their ability
to find an optimum in huge search spaces. In fact, genetic algorithms are
efficient only in systems with very large search spaces. Among the disadvantages
of genetic algorithms is their high demand for computational power. As
a consequence of the high number of necessary evaluations of the fitness
function, each single evaluation has to be cheap in terms of efforts to
obtain the response value.
|
|||||||||
Home Multivariate Data Optimization Survey of Methods Genetic Algorithms |