You are on page 1of 13

Genetic and Evolutionary Algorithms

Kevin Brown -Data Mining Methods-

Presentation Outline

Discuss Evolutionary Algorithms Brief History of Genetic Algorithms Discuss Genetic Algorithms and its Process View Pseudo-code View Sources Q and A

Evolutionary Algorithms

Pertains to Artificial Intelligence Metaheuristic optimization algorithm Subclass of Evolutionary Computation Most popular EA is the Genetic Algorithm (GA)

GA History

It all began with Nils Aall Carricelli (1954) Biologists run amock with GA in the 60s Methods were published in the early 70s by Fraser, Burnell, and Crosby Jon Holland brings GA to the spotlight with his work of the mid 70s

Introduction to Genetic Algorithms

GA search technique used to find solutions to optimization or search problems Categorized as a Global Search Heuristic A Class of EA that use techniques inspired by evolutionary biology Applications include: comp sci., engineering, mathematics, physics, and economics

GA Procedure

Population of individual solutions created Each individual evaluated Most fit are selected The selected are then regrouped New Population is formed Next algorithm iteration begins

GA Initialization

Population of solutions randomly generated Typically very large Used to cover entire search range Occasionally range is optimized Discuss knapsack example

GA Selection

Individuals are selected to reproduce Fitness function weeds out the weak The strong survive to reproduce Poor or weak solutions ruled out FF is stochastic

GA Reproduction

A pair of parents selected Parents create a child solution Child shares attributes with parents Process repeats for generations Solutions evolve End population much different from the first.

GA Termination

Solution is satisfactory Manual evaluation of results Limited number of generations are filled

Pseudo-code

(provided by Wikipedia.org)

Choose initial population Evaluate the individual fitnesses of a certain proportion of the population Repeat: -Select best-ranking individuals to reproduce -Breed new generation through crossover and mutation (genetic operations) and give birth to children -Evaluate the individual fitnesses of the children population -Replace best-ranking individuals Until terminating condition

GA Restrictions

Cannot handle dynamic data Convergence on optima dependant on fitness function Cannot solve yes/no right/wrong problems very well In certain cases simpler algorithms are better than GA GA produces good results in complex data sets

Sources

Genetic Algorithms: http://en.wikipedia.org/wiki/Genetic_algor ithm Genetic Algorithms: Genetic Algorithms in Search, Optimization, and Machine Learning by David E. Goldberg Evolutionary Algorithms: http://en.wikipedia.org/wiki/Evolutionary_ algorithms

You might also like