Schema-based Diversification in Genetic Programming

In genetic programming (GP), population diversity represents a key aspect of evolutionary search and a major factor in algorithm performance. In this paper we propose a new schema-based approach for observing and steering the loss of diversity in GP populations. We employ a well-known hyperschema definition from the literature to generate tree structural templates from the population's genealogy, and use them to guide the search via localized mutation within groups of individuals matching the same schema. The approach depends only on genealogy information and is easily integrated with existing GP variants. We demonstrate its potential in combination with Offspring Selection GP (OSGP) on a series of symbolic regression benchmark problems where our algorithmic variant called OSGP-S obtains superior results.