School of
Information Technology and Electrical Engineering

Spiking Neural Networks with temporal delays enable clock-free learning: how timing representation can be used in forming construals

Speaker: 
Ting Ting (Amy) GibsonTue, 24/09/2013 - 10:00
Venue: 
49-502
Host: 
Prof Janet Wiles
Abstract: 

Forming construals is the process of interpreting and conceptualising the objective world; an important cognitive ability that is used to form judgements and make predictions. This project proposes the development of a cognitive architecture, featuring a bio-inspired spiking neural network, which models this behaviour. Real world events occur sequentially featuring multiple time-scale structures and thus timing representation and the ability to abstract long and varied temporal dependencies are key factors in producing effective construals from time series. The proposed model utilises intrinsic “clock-free” timing representations, combining both state-based and dynamics-based models, to encode the complex temporal dynamics required for forming construals.

 

This talk will present an overview of the cognitive ability of forming construals from time series, and reviews the literature which relates to the aspects of computational modelling that influences the effectiveness of construing. Initial progress of a spiking neural network model that can construe simple textual sequences will be discussed along with a detailed plan for future work.

Biography: 

Amy obtained her Bachelor of Science degree in Computer Science from the University of Queensland and she is currently a 1st year PhD student with the research group Complex and Intelligent Systems in the School of ITEE.  Her main research interests are modelling cognitive behaviours with spiking neural network and analysing real data of brain recordings.

Seminar Type: 

PhD Confirmation Seminar

Pages