Learning on a Continuum in Evolved Dynamical Node Networks

In artificial life, there has been much previous research using evolution to generate learning behaviour within dynamical system controllers without pre-defining the learning mechanisms; so far this research has focused exclusively on evolving agents that can behave differently in a discrete number of different scenarios, generally two. But many (arguably most) interesting discrimination tasks in real life are where the scenarios are over a continuum; one example would be parental imprinting in birds. Here we analyse a successfully evolved embodied and situated agent on an abstract model of this imprinting and give the first published example of such learning on a continuum.