[Paper Review] Gene Expression Programming: a New Adaptive Algorithm for Solving Problems

Cândida Ferreira|ArXiv.org|Feb 25, 2001

Evolutionary Algorithms and Applications20 references2,016 citations

TL;DR

Introduces gene expression programming (GEP), a genotype/phenotype genetic algorithm that evolves linear genomes which encode expression trees, enabling efficient program creation and solving diverse problems.

ABSTRACT

Gene expression programming, a genotype/phenotype genetic algorithm (linear and ramified), is presented here for the first time as a new technique for the creation of computer programs. Gene expression programming uses character linear chromosomes composed of genes structurally organized in a head and a tail. The chromosomes function as a genome and are subjected to modification by means of mutation, transposition, root transposition, gene transposition, gene recombination, and one- and two-point recombination. The chromosomes encode expression trees which are the object of selection. The creation of these separate entities (genome and expression tree) with distinct functions allows the algorithm to perform with high efficiency that greatly surpasses existing adaptive techniques. The suite of problems chosen to illustrate the power and versatility of gene expression programming includes symbolic regression, sequence induction with and without constant creation, block stacking, cellular automata rules for the density-classification problem, and two problems of boolean concept learning: the 11-multiplexer and the GP rule problem.

Motivation & Objective

• Introduce gene expression programming as a new technique for automatic computer program creation.
• Demonstrate how separating genome and expression tree enhances search efficiency compared to existing adaptive methods.
• Show versatility of GEP across problems such as symbolic regression, sequence induction, block stacking, cellular automata, and boolean learning.

Proposed method

• Represent chromosomes as linear genes organized with a head and tail.
• Encode expression trees from chromosomes which are subjected to mutation, transposition, and recombination operators.
• Use a separate genome and expression tree to enable efficient evolution and robust selection of programs.
• Apply genetic operators including one- and two-point recombination to modify chromosomes.

Experimental results

Research questions

• RQ1 How can a genotype/phenotype separation improve the efficiency of evolving computer programs?
• RQ2 Can GEP effectively solve a range of problem types, including symbolic regression, sequence induction, and boolean learning?
• RQ3 What genetic operators best facilitate the evolution of accurate expression trees from linear chromosomes?

Key findings

• GEP encodes expression trees from linear chromosomes with a head and tail.
• The approach uses mutations and various transpositions and recombinations to explore the search space.
• The genome/tree separation yields high efficiency that surpasses existing adaptive techniques.
• GEP is demonstrated on problems including symbolic regression, sequence induction, block stacking, cellular automata rules for density-classification, and 11-multiplexer/GP rule problems.