Representation of input signals in recurrently connected neuronal networks

Events — By Jennifer Thomas on January 10, 2012 11:41 am

Dr Matthieu Gilson, postdoctoral researcher from the world-renowned computational neuroscience RIKEN Brain Science Institute research labs in Saitama, Japan, will present the first Melbourne School of Engineering Visiting Scholar Lecture in 2012 on Wednesday 18 January, at the Centre for Neural Engineering, 203 Bouverie Street, Carlton. The lecture entitled “Representation of input signals in recurrently connected neural networks” will commence at 11.00am, with morning tea served at 10.30am.

Spike-timing-dependent plasticity (STDP) has been observed in many brain areas, such as sensory cortices, where it is hypothesised to develop the structure of synaptic connections between neurons. Previous studies have demonstrated how STDP can capture spiking information at a short timescale, such as coincident spiking, spike patterns and oscillatory spike trains.

It is not yet clear how STDP stores this information in detail in the synaptic structure of recurrently connected neuronal networks. Dr Gilson will use a theoretical framework to investigate the learning dynamics induced by STDP based on the temporal cross-correlograms between input spike trains. The key to understand the computation scheme is the interplay between the properties of STDP, neuronal response and input correlograms.

In the case of a single neuron, STDP can extract the strongest spectral components of the input correlation structure in a similar fashion to principal component analysis (PCA). In a recurrent network, STDP can encode different features of input stimuli relying on partial connectivity and heterogeneities in the synaptic connections (e.g., conduction delays).

This mathematical study of the STDP dynamics aims to bridge the gap between physiology and machine learning, thus shedding light on possible neuronal encoding schemes. On the application side, such spiking neuronal networks that behave as self-adapting filters can be used for the representation, categorisation and detection of spatio-temporal stimuli.

Dr Matthieu Gilson graduated from the Ecole Polytechnique (Palaiseau, France) in 2001. From 2001 to 2004, he completed Masters by Research at the Ecole Polytechnique of Montreal (Canada), where he started to work on neuronal networks. In 2005, he moved to the University of Melbourne and started a PhD at the Department of Electrical and Electronic Engineering, which he completed in 2009. In 2010, he joined the NCT Lab at the RIKEN Brain Science Institute, led by Tomoki Fukai, as a Researcher. He is also an Honorary Research Fellow at the University of Melbourne and has an active collaboration with Professor Anthony Burkitt and Associate Professor Prof David Grayden. His current research interests are mathematical modelling of spike-timing-dependent plasticity (STDP) and data analysis of multiunit neuronal activity recorded in the hippocampus of behaving rats.

For more information about this seminar contact Associate Professor David Grayden in the Department of Electrical and Electronic Engineering.

Tags: biomedical engineering, electrical and electronic engineering, MSE