OMEGA Ordering Messy GA: Solving Permutation Problems with the Fast Messy Genetic Algorithm and Rand

1
OMEGA Ordering Messy GA Solving Permutation
Problems with the Fast Messy Genetic Algorithm
and Random Keys

• Authors Dimitri Knjazew and David E. Goldberg
• Presented by Jian-Hung Chen
• Feng Chia University, Taiwan

2
Outline

• Introduction
• Ordering Messy GA (OmeGA)
• Ordering Deceptive Problems and Problem Coding
• Results
• Conclusions

3
Introduction

• There are many permutation problems in real-world
  applications.
• This paper develops the ordering messy GA (OmeGA)
  specialized for permutation problems.

4
Ordering Messy GA (OmeGA) (1/3)

• OmeGA is based on the fast messy GA (fmGA), all
  basic mechanisms of fmGA are applied.
• Two major modifications in OmeGA
• Messy genes
• Multiple epochs

5
Ordering Messy GA (OmeGA) (2/3)

• Messy genes
• fmGA messy genes (gene locus, gene value)
• OmeGA messy genes (gene locus, random key)
• What is random keys ?
• Real random numbers are used as sort keys to
  decode a sequence.
• A (0.46, 0.91, 0.33, 0.75, 0.51) (3 1 5 4 2)

6
Ordering Messy GA (OmeGA) (3/3)

• Multiple epochs
• Apply fmGA iteratively.
• Best individual found so far is used as a
  competitive template for the succeeding epoch.

7
Ordering Deceptive Problems and Problem Coding

• Two deceptive permutation problems is used to
  test the performance of OmeGA.
• Relative ordering function.
• Absolute ordering function.
• Two kinds of problem coding scheme is used to
  compare the scale-up properties.
• Deflen6 coding defining length six.
• Loss coding the sum of all defining lengths is
  maximal.

8
Results (1/2)

• Deflen6 coding OmeGA and RKGA found global
  optimum in every run.

OmeGA
9
Results (2/2)

• Loose coding RKGA cannot found global solution.

OmeGA
10
Conclusions

• OmeGA is linkage-friendly and coding independent.
• Both relative and absolute ordering deceptive
  problems are solved optimally.