dr Tomasz D.Gwiazda
 Assistant Professor

 ———————
Home page
Short CV
    Publications
   
(e-)Books
   
Papers
My latest book
Students
Office hours
Teaching
 ———————

Contents of e-Book
Index of authors
Index of experiment domains


Introduction

Standard operators
1-Point Crossover
k-Point Crossover
Shuffle Crossover
Reduced Surrogate Crossover
Uniform Crossover
Highly Disruptive Crossover,Heuristic Uniform Crossover
Average Crossover
Discrete Crossover
Flat Crossover
Heuristic Crossover,Intermediate Crossover
Blend Crossover


Binary coded operators
Random Respectful Crossover
Masked Crossover
1bit Adaptation Crossover
Multivariate Crossover
Homologous Crossover
Count-preserving Crossover
Elitist Crossover
    Homologous Crossover  
         

 

 

(HX)

download PDF with first 40 pages
from my latest eBook

if you need more operators click here

Keywords
information exchanging
, information destruction, convergence speed, non-disruptive crossover, optimal crossover points

Motivation
   Reduction of the destructive action of a multi-point crossover operator resulting from random selection of crossover points

Source text
   Park J., Park J., Lee Ch., Han M. (1993), Robust and Efficient Genetic Crossover Operator: Homologous Recombination, in Proceedings of 1993 International Joint Conference on Neural Networks, pp. 2975-2978
WEB:      http://ieeexplore.ieee.org/xpl/abs_free.jsp?arNumber=714347

Read also
   Beyer H.-G. (1995), Toward a Theory of Evolution Strategies: On the Benefits of Sex – the (μ/μ,lambda) Theory, in Evolutionary Computation, vol. 3(1), pp. 81-111
WEB:     http://citeseer.ifi.unizh.ch/361714.html
            http://citeseer.ist.psu.edu/361714.html

See also
   Spontaneous Crossover
   Circle-ring Crossover
   Sufficient Exchanging
   Intermediate Crossover
   Intermediate Crossover2
   Adaptive Number of Crossover Points

Algorithm
1.
     select two parents A(t) and B(t) from a current population P(t)

2.     randomly choose m crossover points {pc1,...,pcm}

3.     create two offspring A(t+1) and B(t+1) by restricted m-point crossover
    as follows:

4.     for every pair of strings STA=(apck(t),...,apck+1(t)) and
   
STB=(bpck(t),...,bpck+1(t)) between two successive crossover points
    pck and pck+1 do:

5.              if length_of_STA (= length_of_STB ) ≥ w then

6.              compute the degree of similarity DS of strings STA  and STB
              
 as follows:

7.              number_of_1 = 0

8.                              for i = k  to k+w do

9.                                             if apci(t) XOR bpci(t) =1 then

10.                                           number_of_1 = number_of_1 + 1

11.                                           end if

12.                           end do

13.            DS = number_of_1/length_of_STA

14.                           if DS ≥ u  then

15.                           swap bits

16.                           else

17.                           do nothing

18.                           end if

19.            else

20.            do nothing

21.            end if

22.  end do

where:
w
and u – parameters of the method

Comments
   The HX operator is based on the standard k-Point Crossover operator. Introduced modification relies on the fact that only strings of bits which are at least of a certain length (row: 5) or of an admissible degree of  similarity (row: 14) are allowed to crossover. Determination of the degree of similarity is based on the XOR operator (rows: 9-13).
   This strategy is aimed at transferring (hence also protecting) strings with specified parameters to the next generation.
   In HX the value of w and u is determined à priori as constant or dynamically changed (increased) in the GA run.

Experiment domains
   De Jong’s function (F1)

Compared to
   k-Point Crossover
   Uniform Crossover
   Intermediate Crossover
   Intermediate Crossover2

 
   

    :: Copyrights © tomaszgwiazda e-books 2006 :: webmaster ::