A non-destructive grammar modification approach to modularity in grammatical evolution

Abstract

Modularity has proven to be an important aspect of evolutionary computation. This work is concerned with discovering and using modules in one form of grammar-based genetic programming, grammatical evolution (GE). Previous work has shown that simply adding modules to GE's grammar has the potential to disrupt fit individuals developed by evolution up to that point. This paper presents a solution to prevent the disturbance in fitness that can come with modifying GE's grammar with previously discovered modules. The results show an increase in performance from a previously examined grammar modification approach and also an increase in performance when compared to standard GE.

Modular Grammatical Evolution for the Generation of Artificial Neural Networks

“…A modular solution contains almost independent units (modules) where each unit plays a specific role in solution's performance (Amer and Maul, 2019). Modularity provides important advantages: reuse, readability, and scalability (Koza, 1994;Swafford et al, 2011a). While the modules may be highly complex internally, they are loosely connected together usually in a modular solution.…”

Natural Computing Algorithms

“…Researchers have explored different approaches to modularity in GE, ranging from static grammar-defined functions (a variation on ADFs) [469] dynamically defined variants using dynamic grammars (i.e., grammars which automatically update to incorporate new ADFs) [255] and metagrammardefined ADFs [263]. More recently, Swafford et al have examined a number of different approaches to identify and then incorporate subderivation trees as modules [608,609,610,611,612,607]. In all of the above approaches, modularity is found to provide performance gains on problems which have sufficient 'difficulty' to warrant the overhead of increasing the search space by including mechanisms for modularity.…”

Automated Problem Decomposition for the Boolean Domain with Genetic Programming

“…Several other approaches for the creation/identification of modules have been proposed in the literature-e.g., in the context of grammar-based GP (grammatical evolution) [10,11], in Cartesian Genetic Programming (CGP) [12] and in GP systems using the Push language [13]; other approaches are discussed in [14]. The majority of approaches for modularity (including the ones discussed above) focus on the discovery of modules rather than on the use of modules to decompose the problem into smaller (more tractable) subproblems, relying on the idea that if modules can be created/identified, their usefulness will emerge through the GP search.…”