Geometric Semantic Genetic Programming

Abstract

The version in the Kent Academic Repository may differ from the final published version. Users are advised to check http://kar.kent.ac.uk for the status of the paper. Users should always cite the published version of record.

Automated Problem Decomposition for the Boolean Domain with Genetic Programming

“…3 Therefore, a natural way of combining the partial solutions is to combine their output vectors. To that end, we use the semantic crossover proposed by Moraglio et al [17] to generate mask selectors, which act as tests to inform which of the partial solutions to use for a given input.…”

“…The geometric semantic crossover [17] is a semantic operator that works on the output vector of two individuals (candidate solutions). For the Boolean domain, the semantic crossover (SGXB) returns an individual…”

“…The construction of the individual T 3 is illustrated in Figure 3. We will focus on the Boolean domain from now on; refer to [17] for details of how to apply the semantic crossover in other domains. Fig.…”

Preliminary Study of Bloat in Genetic Programming with Behavior-Based Search

“…Semantics in GP describes the performance of a program with the raw output vector computed over all fitness cases [21,1,2,12,20,36]. Given a set of n fitness cases, the semantics of a program K is the corresponding output vector it produces y ∈ R n .…”

“…Therefore, semantics adds another space of analysis in which the search is being conducted, along with genotypic, phenotypic and objective (fitness) space, we can also consider semantic space; where a many-to-one relation will usually exists between genotypic (phenotypic) space and semantic space. Researchers have used semantics to improve GP in different ways, such as modifying traditional genetic operators to improve the semantic diversity of the evolving population [1,2,36], or by explicitly performing evolution within semantic space [12,21]. In general, all of these works have shown improved results using a canonical GP as a control method, mostly on symbolic regression problems.…”