Skip navigation
PDF 1.16 MB
Pages 916-923
First published 30 July 2014

Evaluating Topological Models of Neuromodulation in Polyworld

Jason Yoder, Larry Yaeger

Abstract (Excerpt)

Neuromodulation or the modification of neural activity can enhance the biological capabilities of organisms for learning and adaptation (Doya, 2002). Computational models of neuromodulation have demonstrated an advantage for specific tasks (McHale and Husbands, 2003; Parussel, 2006; Soltoggio et al., 2008). In this paper we introduce simple, topological models of neuromodulation and provide a preliminary evaluation of their performance in a foraging task implemented in Polyworld. This work is novel in its evaluation of neuromodulatory models on tasks involving multiple agents competing in an ecologically demanding environment. Polyworld abstracts away the complexity of basic locomotion and digestion, but is biologically consistent in energy conservation, and simulations routinely involve populations of hundreds of neurally controlled agents. The work is also novel in its evaluation of a diverse mixture of models, employing two distinct neural growth models and two distinct models of neuromodulation to draw distinctions otherwise difficult to ascertain. The models developed in Polyworld are explained, evaluated and compared with other computational models of neuromodulation. In particular we find that neuromodulation may be able to enhance or diminish foraging performance in a competitive, dynamic environment, depending upon the nature of the action of neuromodulation and the distribution of neuromodulatory sources.