Evolving neuromodulator architectures on non-associative learning tasks
Jason A. Yoder · 2017
Neuromodulation is an integral process in neural systems. Even the most well-studied neural circuits and organisms cannot be understood effectively without taking into account the effects of neuromodulators. The term neuromodulation refers to a broad range of phenomena, each of which has important computational distinctions. In this paper, we report on results measuring performance of differing forms of computational neuromodulation on a biologically meaningful, non-associative learning task. A novel neuroevolution approach, GasNEAT, is introduced and used to conduct the experiments. Findings indicate that, under certain conditions, networks with neuromodulation are more likely to evolve habituating behaviors than those without.