A Fast Approach for Cross over of two tree-shape ADTs in genetic programming

Vahid Haghighatdoost | Maryam Espandar | Alireza Bossaghzadeh

---

URL :   http://research.shahed.ac.ir/WSR/WebPages/Report/PaperView.aspx?PaperID=3144
Date :  2010/06/29
Publish in :    Knowledge Generation, Communication and Management


Keywords :Cross

Abstract :

---

This paper presents a new method for crossover of two tree-shape chromosomes in a genetic algorithm to solve problems that their populations are a set of tree ADTs1. Each tree is a solution of the target problem and in training steps we are looking for the best tree. All Trees of our population have the same nodes with different connections, in other words they are different in their topologies. The cross over operates on two parent chromosomes and generates two new chromosomes as children. Selection of links from parents without any methodology will not generate a tree, because of the new child would have cycles or consist of separate parts. The proposed algorithm for crossover of two trees which is called C_T generates two new children from two input tree structured-chromosomes