Access Type

Open Access Dissertation

Date of Award

January 2015

Degree Type

Dissertation

Degree Name

Ph.D.

Department

Computer Science

First Advisor

Robert G. Reynolds

Abstract

Cultural Algorithms are computational models of social evolution based upon principle of Cultural Evolution. A Cultural Algorithm are composed of a Belief Space consisting of a network of active and passive knowledge sources and a Population Space of agents. The agents are connected via a social fabric over which information used in agent problem solving is passed. The knowledge sources in the Belief Space compete with each other in order to influence the decision making of agents in the Population Space. Likewise, the problem solving experiences of agents in the Population Space are sent back to the Belief Space and used to update the knowledge sources there. It is a dual inheritance system in which both the Population and Belief spaces evolve in parallel over generations.

A question of interest to those studying the emergence of social systems is the extent to which their organizational structure reflects the structures of the problems that are presented to them. In a recent study [Reynolds, Che, and Ali, 2010] used Cultural Algorithms as a framework in which to empirically address this and related questions. There, a problem generator based upon Langton's model of complexity was used to produce multi-dimensional real-valued problem landscapes of varying complexities. Various homogeneous social networks were then tested against the range of problems to see whether certain homogeneous networks were better at distributing problem solving knowledge from the Belief Space to individuals in the population. The experiments suggested that different network structures worked better in the distribution of knowledge for some optimization problems than others. If this is the case, then in a situation where several different problems are presented to a group, they may wish to utilize more than one network to solve them. In this thesis, we first investigate the advantages of utilizing a heterogeneous network over a suite of different problems. We show that heterogeneous approaches begin to dominate homogeneous ones as the problem complexity increases. A second heterogeneous approach, sub-culutres, will be introduced by dividing the social fabric into smaller networks.

The three different social fabrics (homogeneous, heterogeneous and Sub-Cultures) were then compared relative to a variety of benchmark landscapes of varying entropy, from static to chaotic. We show that as the number of independent processes that are involved in the production of a landscape increases, the more advantageous subcultures are in directing the population to a solution. We will support our results with t-test statistics and social fabric metrics performance analysis.

Share

COinS