The thesis addresses the most fundamental architectural characteristic of the brain: the fact that it is divided into two halves, with resulting functional asymmetries. The goal is to explore the characteristic behaviours, and the advantages if any, of processing in a split architecture. The explorations are concerned with the fundamental issue of how information is handled and distributed between different parts of a whole system. Specifically, the thesis is concerned with the intersection of visual processing in humans, marked by the separation of the visual cortices, and the resource allocation properties of split neural networks. This study involves<sub> </sub>the representational differentiation and associated behaviours that result from the interaction of several factors, and particularly: the developing architecture; the way in which stimuli are presented to the processor; and the nature of the stimuli themselves. I present results from connectionist modelling experiments with split architectures. The principal issues addressed in the thesis revolve around what the advantages might be for bi-hemispheric processing of visual information, and how real-time high-density information management---such as that employed in the human visual system---copes with the fact that processing of the same stimulus is done in two halves, in two different places. Despite the brain's clear division into two cerebral hemispheres, there has to date been relatively little computational modelling of this aspect of gross brain morphology. However, certain recent models employing split neural networks have demonstrated that such architectures are not only effective for processing visual information, but that certain emergent processing strategies are particular to these split architectures. This thesis reports further studies of the complex relationship that comes into play between particular architectural features and general processing strategies, as well as distinct variations in the nature of the stimuli involved. The particular feature of the visual system addressed in this thesis is the recently acknowledged vertical splitting of the human retina, including the fovea, resulting in each visual hemifield initially being precisely projected to the contralateral hemisphere. As an example of the emergent effects that models of such a system produce, this thesis examines the nature of symmetrical forms, asking in what ways the perception of symmetry can be treated as a special case in cognition. This thesis makes a contribution to our understanding of processing in a divided architecture such as the human brain, and to unpacking the causes that lie behind functional lateralisation, by examining split processing in its most general form.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:652417 |
| Date | January 2002 |
| Creators | Hicks, John William James |
| Publisher | University of Edinburgh |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://hdl.handle.net/1842/12080 |
Page generated in 0.002 seconds