In genetic algorithms, if two parents are two graphs or two trees, instead of two sequences, then how to design the crossover operator?
See what Google says "randomly swap it, or compose it out of two parts of two parental (sub)trees", where it = tree/ subtree.
Perhaps ChatGPT would provide you with even smarter guesses.
Technically, Fig. 3 and 4, next, show crossover and mutation.
Article Optimization of Combinational Logic Circuits with Genetic Programming
One approach could be to chose one random vertex in each parent graph. Then swap corresponding sub trees.
Second approach could be to somehow represent graph in a linear way (array). Than crossover is natural.
Hope that this helps.
I suggest you to read any paper on genetic programming, where solutions are mathematical formulae coded as trees
You can use a binary array to hold information about open and closed links between nodes, as shown in the image. The tree is built by the closed links (value "1").
You can do crossover by swapping parts of the "parent" arrays.
However, you will need an algorithm to fix those "child" arrays that have isolated nodes or that have loops.
DR.Li
You can use Multi-gene genetic programming where more than two parents are involved in the recombination operation.
Conference Paper Genetic algorithms with multi-parent recombination
GOOD LUCK
Dear Hui,
I proposed a subgraph crossover for Cartesian Genetic Programming in 2017:
Conference Paper A New Subgraph Crossover for Cartesian Genetic Programming
Last year, I proposed discrete recombination for Cartesian Genetic Programming:
Chapter Towards Discrete Phenotypic Recombination in Cartesian Genet...
Cartesian Genetic Programming is a graph-based Genetic Programming variant.
I hope that this is helpful for your work
Kind regards
Roman
- Badges
- Science topic
- Similar topics
- Biological Science
- Histology
- Sectioning