Genetic Programing - Embedded Intelligent Robotics - Lecture Slides, Slides of Robotics

Download Genetic Programing - Embedded Intelligent Robotics - Lecture Slides and more Slides Robotics in PDF only on Docsity!

Agenda

• (^) Future of Genetic Programming.• (^) Examples of Genetic Programs.• (^) How Genetic Principles are Applied.• (^) Why Genetic Programming?• (^) Background/History. • (^) What is Genetic Programming?

Genetic AlgorithmsGenetic Algorithms

can be evaluatedof each solutionquality or fitnessbefore thephenotypesbe mapped to Genotypes must – solution space - actual solutions (phenotypes)– search space - coded solution (genotype) • (^) uses 2 separate spaces• (^) Robust • (^) Most widely used

• Individuals are represented asspace • Like GP no distinction between search and solution Evolutionary StrategiesEvolutionary Strategies

(^) real-valued (^) vectors.

• (^) Susceptible toproblem parameters of the parent.– Child solution generated by randomly mutating the– one parent and one child• (^) Simple ES (^) stagnation at local optima

Genotype = Phenotype directly (ie^ Genotype = Phenotype •^ Like ES operates on the decision variable of the problemonly•^ One population of solutions, reproduction is by mutationtransition table of finite state machines•^ Early versions of EP applied to the evolution of^ •^ Resembles ES, developed independently Evolutionary ProgrammingEvolutionary Programming )

fitness deleted.– everyone evaluated and the half of population with lowest– children generated until population doubled in size– better fitness more likely a parent•^ Tournament selection of parents

ocsity.co AlgorithmsAlgorithmsEvolutionaryEvolutionaryArchitecture ofArchitecture ofGeneralGeneral

Background/History

evolved tree structures.• Combining the idea of machine learning andSelection.” - Origin of GP.of Computers by Means of Natural“Genetic Programming. On the Programming• 1992, Genetic Programming Treatise -^ • By John R. Koza, Stanford University.

• Again, Artificial Intelligence.creating ones that give optimal solutions.• Not only designing the algorithms buthaving to design complex algorithms. • It saves time by freeing the human from Why Genetic Programming?

• Fitness testing.• Mutations.• Crossovers. • “Breeding” computer programs. How are Genetic Principles Applied?

• (^) (2 + a)*(4 - num) • (^) Infix/Postfix Computer Programs as Trees (^2) a 4 + (^) - *

ocsity.co num

The Fitness Test

distance, time, etc…• Can be measured in many ways, i.e. error,environment.• How good can a program cope with itsproblem at hand.given computer program is at solving the^ • Identifying the way of evaluating how good a

Fitness Test Criteria

• i.e. n 2 • Time complexity a good criteria.

vs. nlogn.

tested.• Combinations of criteria may also be• Accuracy - Values of variables.

Mutations in Programs

• (^) Single parental program is (^) probabilistically selected (^) from
• a (^) new subtree is grown – the subtree rooted at that point is deleted, and• Mutation •^ Mutation (^) point randomly chosen.the population based on fitness. (^) there using the same random growth
• (^) Asexual operations process that was used to generate the initial population. (^) (mutation) are typically performed
• with a (^) low probability sparingly: (^) of,
• (^) probabilistically selected from the population based on fitness.

new genes created •^ No^ new genes created between two previous chromosomes.•^ No loss of genes, but an exchange of geneshybrid combinations (recombinant).their genetic information to create new^ •^ Two parental chromosomes exchange part of Crossovers in Nature

, preexisting old ones

mixed together.