Towards a Biologically Plausible Active Visual Search Model

This paper proposes a neuronal-based solution to active visual search, that is, visual search for a given target in displays that are too large in spatial extent to be inspected covertly. Recent experimental data from behaving, fixating monkeys is used as a guide and this is the first model to incorporate such data. The strategy presented here includes novel components such as a representation of saccade history and of peripheral targets that is computed in an entirely separate stream from foveal attention. Although this presentation describes the prototype of this model and much work remains, preliminary results obtained from its implementation seem consistent with the behaviour exhibited in humans and macaque monkeys.