Evolutionary Algorithms Principles, Methods and Algorithms

Evolutionary Algorithms Principles, Methods and Algorithms

When to use genetic algorithms:

When not much is known about the response surface and computing the gradient is either computationally intensive or numerically unstable many scientists prefer to use optimization methods such as genetic algorithms, simulated annealing, and Simplex optimization which do not require gradient information. One of the reasons that scientists prefer to use genetic algorithms is their versatility. Using knowledge about the system we can tailor the algorithm for a particular application. If the application calls for an optimization method with hill-climbing characteristics the algorithm can be modified by using an elitist strategy. If becoming trapped in local optima is a problem, mutation can be increased. Thus, while there is no guarantee that GA will perform the best for a particular application we can usually change some aspect of the genetic configuration or use different genetic operator to achieve adequate search performance.
Another feature about GA that we take advantage of is that GA do not optimize directly on the variables but on their representations. For SUB-type applications such as wavelength selection having the chromosome coded such that each gene represents a particular wavelength or group of wavelengths is particularly useful.
GA is an excellent choice for these types of applications, where the variables being optimized are very different from each other (i.e. a mixture of integers, binary values, and floating points numbers)? The GA configuration can be modified to include different mutation operators for different sections of the chromosome.


for more info visit.
http://www.enjineer.com