Computational Models of Conditioning: Evolution of attention in learning

A variety of phenomena in associative learning suggest that people and some animals are able to learn how to allocate attention across cues. Models of attentional learning are motivated by the need to account for these phenomena. We start with a different, more general motivation for learners, namely, the need to learn quickly. Using simulated evolution, with adaptive fitness measured as overall accuracy during a lifetime of learning, we show that evolution converges to architectures that incorporate attentional learning. We describe the specific training environments that encourage this evolutionary trajectory, and we describe how we assess attentional learning in the evolved learners.