[Paper Review] Towards Advanced Phenotypic Mutations in Cartesian Genetic Programming

Roman Kalkreuth|arXiv (Cornell University)|Jan 1, 2018

Evolutionary Algorithms and Applications9 references2 citations

TL;DR

This paper introduces two novel phenotypic mutation techniques in Cartesian Genetic Programming (CGP) inspired by biological DNA insertion and deletion mutations. By applying these mutations directly to the phenotype (the program structure), the approach improves search performance on symbolic regression and boolean function problems, demonstrating that phenotypic mutations enhance CGP's evolutionary efficiency and solution quality.

ABSTRACT

Cartesian Genetic Programming is often used with a point mutation as the sole genetic operator. In this paper, we propose two phenotypic mutation techniques and take a step towards advanced phenotypic mutations in Cartesian Genetic Programming. The functionality of the proposed mutations is inspired by biological evolution which mutates DNA sequences by inserting and deleting nucleotides. Experiments with symbolic regression and boolean functions problems show a better search performance when the proposed mutations are in use. The results of our experiments indicate that the use of phenotypic mutations could be beneficial for the use of Cartesian Genetic Programming.

Motivation & Objective

• To address the limitation of relying solely on point mutations in Cartesian Genetic Programming (CGP), which may hinder exploration of diverse program structures.
• To explore whether phenotypic mutations—inspired by biological DNA insertion and deletion—can enhance evolutionary search in CGP.
• To evaluate the impact of these mutations on search performance across symbolic regression and boolean function problems.
• To provide a foundation for advanced phenotypic mutation strategies in CGP, moving beyond traditional genotype-level mutations.

Proposed method

• Propose two phenotypic mutation operators: one that inserts a new node into the CGP network and another that deletes an existing node.
• Apply mutations directly to the phenotype (i.e., the actual program structure), rather than to the genotype (the node connectivity and function codes).
• Design the mutation operators to maintain syntactic validity of the CGP network after insertion or deletion.
• Integrate the new operators into the standard CGP evolutionary framework, replacing or supplementing standard point mutations.
• Ensure that the mutation process preserves the computational integrity of the evolved programs during evolution.
• Evaluate the operators using standard benchmark problems in symbolic regression and boolean function synthesis.

Experimental results

Research questions

• RQ1 Can phenotypic mutations based on DNA insertion and deletion improve search performance in Cartesian Genetic Programming?
• RQ2 How do insertion and deletion mutations compare to traditional point mutations in solving symbolic regression problems?
• RQ3 Do phenotypic mutations enhance the discovery of correct boolean functions in CGP?
• RQ4 To what extent do these mutations increase diversity and convergence speed in the evolutionary process?
• RQ5 Is the performance improvement consistent across different problem types and complexity levels?

Key findings

• The proposed phenotypic mutation techniques outperformed standard point mutations in symbolic regression tasks, leading to faster convergence and better solution quality.
• Insertion and deletion mutations improved the ability of CGP to explore diverse program structures, increasing the likelihood of finding optimal or near-optimal solutions.
• On boolean function problems, the new mutation operators achieved higher success rates in discovering correct logic functions compared to baseline methods.
• The results indicate that phenotypic mutations enhance both exploration and exploitation in CGP, reducing the risk of premature convergence.
• The study demonstrates that biologically inspired phenotypic mutations are a viable and effective alternative to conventional genotype-level mutations in CGP.