Lecture_4 - iologically Inspired Methods (7CEMM708)...

Info iconThis preview shows pages 1–8. Sign up to view the full content.

View Full Document Right Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: iologically Inspired Methods (7CEMM708) iologically Inspired Methods (7CEMM708) Dr. Lei Ren Division of Engineering, Kings College London, Strand, London, WC2R 2LS E-mail: lei.ren@kcl.ac.uk Room 244, Strand Building, Strand Campus Genetic Algorithm: Genetic Algorithm: An Added Level of Sophistication An Added Level of Sophistication (I) (I) Hybrid GA Parallel GA Introduction Introduction A: Added Level of Sophistication (I) A: Added Level of Sophistication (I) Initial Population Initial Population Rating Rating Selection Selection Reproduction Reproduction Mutation Mutation Variable & Cost Function Variable & Cost Function Convergence Convergence Expensive cost functions Multi-objective optimisation Gary code & Gene size Permutation problems Population initialization Alternative crossover Alternative mutation Convergence check Expensive Cost Functions Expensive Cost Functions A: Added Level of Sophistication (I) A: Added Level of Sophistication (I) Handling the Expensive Cost Functions The cost function is sometimes extremely complicated and time-consuming to evaluate. To reduce the number of function evaluation 1. I nitial population is created with no two identical chromosomes Generally, only a problem for binary GA Checking the population for repetitions is time-consuming. An efficient approach: begin each chromosome with a different pattern Expensive Cost Functions Expensive Cost Functions A: Added Level of Sophistication (I) A: Added Level of Sophistication (I) Handling the Expensive Cost Functions To reduce the number of function evaluation 1. I nitial population is created with no two identical chromosomes Expensive Cost Functions Expensive Cost Functions A: Added Level of Sophistication (I) A: Added Level of Sophistication (I) Handling the Expensive Cost Functions To reduce the number of function evaluation 2. To generate population with all unique members a. New offspring or mutated chromosomes are checked against the chromosomes in the current population, and discard duplication. b. Not allow identical chromosomes to mate c. Keep track of all chromosomes and costs over all the generations. The cost of twin is assigned according to master list. Reserved only for the most difficult cost functions. Expensive Cost Functions Expensive Cost Functions A: Added Level of Sophistication (I) A: Added Level of Sophistication (I) Handling the Expensive Cost Functions Simplify the cost function for the bulk of the runs Define a simpler or lower order cost function in the initial generations Increase the cost function accuracy as GA goes through the generations Use a coarse grid GA to find the valley of the global minimum then use a fast local optimizer to find the bottom of the valley (Hybrid GA) The method does not use more time for searching than it saves in cost function evaluations. Multiple Objective Optimisation Multiple Objective Optimisation A: Added Level of Sophistication (I) A: Added Level of Sophistication (I) Multiple Objective Optimisation...
View Full Document

This note was uploaded on 03/27/2010 for the course MSC ADVANCE SO taught by Professor Dr.markhurman during the Spring '09 term at King's College London.

Page1 / 39

Lecture_4 - iologically Inspired Methods (7CEMM708)...

This preview shows document pages 1 - 8. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online