Analytic Programming – Symbolic Regression by Means of Arbitrary Evolutionary Algorithms

This contribution introduces analytical programming, a novel method that allows solving various problems from the symbolic regression domain. Symbolic regression was first proposed by J. R. Koza in his genetic programming and by C. Ryan in grammatical evolution. This contribution explains the main principles of analytic programming, and demonstrates its ability to synthesize suitable solutions, called programs. It is then compared in its structure with genetic programming and grammatical evolution. After theoretical part, a comparative study concerned with Boolean k-symmetry and k-even problems from Koza’s genetic programming domain is done with analytical programming. Here, two evolutionary algorithms are used with analytical programming: differential evolution and self-organizing migrating algorithm. Boolean k-symmetry and k-even problems comparative study here are continuation of previous comparative studies done by analytic programming in the past.