Structural diversity for genetic algorithms and its use for creating individuals
Masaki Suzuki, Setsuo Tsuruta, Rainer Knauf · 2013
The paper presents a structural representation of diversity (respectively similarity). This representation can be used to decide, whether or not a new individual meets a requested degree of diversity, but also to estimate and optimize a population's coverage of the solution space to avoid running into a local optimum and missing the global one. Moreover, it can also be constructively used for systematically creating new individuals, which (1) meet a certain diversity requirement, (2) additionally improve the coverage of the solution search space, and (3) have an optimal fitness value.