Computational modelling of hand-centred visual representations in the primate brain

Thesis

Neurons have been found in various areas of the primate brain that respond to the location of objects with respect to the hand. In this thesis we present a number of self-organising theories of how these hand-centred visual receptive fields could develop during visually guided learning in unsupervised neural network models. Experimental chapters 3 and 4 show the development of hand-centred representations under two different learning hypotheses Continuous Transformation learning and Trace learning, using a neural network model of the primate visual system, VisNet. In Chapter 5 the trace learning hypothesis is advanced by gradually increasing the realism in the visual training conditions. Results are shown using realistic visual scenes consisting of multiple targets presented at the same time around the hand. In Chapter 6 this learning hypothesis was further explored using natural eye and head movements recorded from human participants. This required a new time accurate differential formulation of the VisNet model to faithfully represent the temporal dynamics of the recorded gaze changes. These results provide an important step in showing how localised, hand-centred receptive fields could emerge under more ecologically realistic training conditions. In Chapter 7 we present a new self-organising model of how hand-centred representations may develop using a proprioceptive signal representing the position of the hand. In these simulations we first develop a body-centred representation of the location of the visual targets by trace learning. I then showed how these body-centred representations of the visual targets can be combined with a proprioceptive representation of the location of the hand in order to develop hand-centred representations. This thesis concludes with a discussion of the main findings and some suggestions for future work.

Authors: Gonzalez, J

• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English
• Keywords: Reference frames; VisNet; Hand-centred; Neural networks
• Subjects: Computational neuroscience; Psychology, Experimental