A multiple expression alignment framework for genetic programming

Abstract

Alignment in the error space is a recent idea to exploit semantic awareness in genetic programming. In a previous contribution, the concepts of optimally aligned and optimally coplanar individuals were introduced, and it was shown that given optimally aligned, or optimally coplanar, individuals, it is possible to construct a globally optimal solution analytically. As a consequence, genetic programming methods, aimed at searching for optimally aligned, or optimally coplanar, individuals were introduced. In this paper, we critically discuss those methods, analyzing their major limitations and we propose new genetic programming systems aimed at overcoming those limitations. The presented experimental results, conducted on four real-life symbolic regression problems, show that the proposed algorithms outperform not only the existing methods based on the concept of alignment in the error space, but also geometric semantic genetic programming and standard genetic programming.