Genetics Algorithms

Genetic algorithms are a part of evolutionary computing, which is a rapidly growing area of artificial intelligence.

This blog covers the canonical genetic algorithm as well as more experimental forms of genetic algorithms, including parallel island models and parallel cellular genetic algorithms. The tutorial also illustrates genetic search by hyper plane sampling. The theoretical foundations of genetic algorithms are reviewed, include the schema theorem as well as recently developed exact models of the canonical genetic algorithm.

There is three common methods that solve a global optimization problems: heuristic, approximation and systematic methods. GA’s are typically heuristic methods. They are based in one hand on a heuristic gradient ascension method (selection & crossover) and, in another hand, on a semi-random exploration method (mutations). Advantages of GA’s are that they are simple to understand and to implement, and early give a good near solution.

Disadvantages are that they tend to fail with the more difficult problems and need good problem knowledge to be tuned.

GA’s are inspired from biological processes (i.e. cells’ division, DNA, crossover, mutation,). The underlining idea is to generate successive sets of solutions (generations), making each new generation inheriting properties from the best solutions of the precedent. In order to perform a step, we have to select the best solutions and mix them together (crossover). A GA typically looks like that:

a) Generate a first generation with random parameters.

b) Evaluate all individuals of the generation.

c) Crossover the best individuals together to get the new generation (children).

d) Make random mutation across the new generation.

e) Go back to b).