Research Interests

Current Research

MULTI-SENSORY INTEGRATION IN BIOLOGICAL NEURAL SYSTEMS AS A STUDY OF INFORMATION PROCESSING

Projects

Understanding diversity of neuronal and circuit processing of multi-sensory inputs from a normative perspective with information theory

Areas involved in multi-sensory integration have neurons with a range of responses from those only responding to a single sensory modality to those whose responses are super-additive on the single sensory responses. This project aims to use information theoretic concepts such as mutual information and Pareto optimality to understand the computations performed by these neurons and formulate a theory as to the collective circuit function.

Understanding how neuronal circuits processing multi-sensory information develop in a stimulus dependent manner

This project aims to use recurrent neural spiking networks with plasticity to investigate how the development of multi-sensory brain areas, such as RL in the mouse, is dependant on stimulus presentation and feedback.

Previous Research

THE OPTIMISATION OF THE DEVELOPMENT AND SPATIAL ORGANISATION OF SENSORY NEURAL NETWORKS

Projects

Understanding evolutionary invariance of orientation preference maps in primates, carnivores, scandentia and primates

This project involves the creation of robust analytical techniques that can be applied to intrinsic imaging data from a wide range of labs and species. This method aims to maximise comparability of results and to understand the origins and extent of noise in different data sets to enhance this comparability. These techniques can then be applied to diverse species such as the Australian fat-tailed dunnart, which we show to be the smallest animal to have organised orientation preference maps. Finally, the current methods of orientation preference map comparison are reliant on singularities of the organisation alone and is independent on columnar variation. We therefore investigate new comparison methods that overcome the under sampling problem inherent in much of biology by constructing and comparing probability mass densities.

Optimisation of orientation preference maps to minimise turnover in neuronal preference during development

Neural turnover during development allows for a system to allow for a self-organising process which can be either dependent or independent of stimulation as required by a system to reach a stable processing state without all features of the end state having to be genetically encoded. Despite the benefits of allowing learning, turnover can pose challenges to systems particularly when the turnover occurs early in a processing pathway and when it has the potential to cause instability in later processing. This is the case for orientation preference maps, one of the first visual processing stages. We investigate this phenomenon numerically and with theoretical models. Numerically we investigate techniques for robust simulation of orientation preference map models while allowing for biologically realistic disorder which we show to have an impact on turnover. Theoretically we construct an optimisation model based on minimising turnover during development and find that turnover is indeed minimal for biologically realistic patterns. We are interested in using the numerical simulations to further understand the nature of turnover and to test predictions of the model such as that changes in any orientation direction is energetically neutral. Such predictions should also be tested experimentally.

Optimisation of the spatial organisation of orienation preference maps for spatial predicitive information

This project involves the extension of Gaussian information bottleneck theory to two dimensions such that it can be applied to the spatial organisation of orientation preference maps. The field representing the spatial location and orientation preference of a neuron is optimised such that it throws out local information about a visual field ensemble and keeps information about distant location in the visual field. We are then able to investigate the field properties for different parameter combinations by deriving a loss function from the third order term and constructing a phase diagram. We are interested in understanding the parameters of the quasi-periodic fields that qualitatively and quantitatively resemble the biological orientation preference maps.