home · mobile · calendar · defenses · 2004-2005 · 

Thesis Defense - Swiercz

Synaptic Activity Plasticity Rule and Its Applications in Networks of Spiking Neurons
Waldemar (Valdi) Swiercz
Computer Science PhD Candidate

An important research focus in neuroscience is directed towards understanding complex interactions occurring between neurons' ensembles within large sections of the brain. This can be partially addressed via computational modeling using artificial neural networks, allowing for example to analyze particular phenomena or hypotheses that are not feasible to check via biological experiments. Furthermore, computational results may lead to new hypotheses and help design new biological experiments. Development of computationally efficient and biologically correct artificial neural network models requires interdisciplinary effort. The computational part requires employment of artificial neuron model, defining synaptic plasticity, and appropriate network stimulation that must be able to work as a stable ensemble, producing reliable results within reasonable period of time. On the other hand, biological experiments need to provide biological data about each element of the network as well as information about how they communicate between themselves. Finally, computational results must be validated via biological experiments.

This research reports the development of an artificial network model that employs an enhanced and optimized MacGregor's integrate-and-fire neuron model, a new synaptic plasticity rule, and the new synapse model with releasable glutamate mechanism. Our model allows for observation of network activity in two-dimensions that is important for analyzing spatial factors that have influence on the network.

The focus of our research is on observing the influence of releasable glutamate, a presynaptic parameter, on the network bursts termination, as well as its impact on long-term potentiation. Understanding the physiology of neurotransmitter release regulation can lead to discover new strategies for the treatment of diseases such as epilepsy.

The network model is based on hippocampal areas CA3 and CA1. The model uses biological results describing spontaneous synchronous bursts of network activity. We validate our model by confirming the results of physiological experiments via computational stimulations. Additionally, we have been able to computationally validate a theory not possible to test physiologically, namely, we showed that inhibitory inputs alone are not able to terminate network bursts. We were also able to determine various ways of controlling spontaneous bursts, for example by altering the supply and release of presynaptic glutamate, or by decreasing the network synaptic potentiation level.

Committee: Krzysztof Cios, University of Colorado at Denver (Chair)
Clayton Lewis, Professor
James Martin, Associate Professor
Michael Mozer, Professor
Diego Restrepo, University of Colorado Health Sciences Center
Kevin Staley, University of Colorado Health Sciences Center
Department of Computer Science
University of Colorado Boulder
Boulder, CO 80309-0430 USA
May 5, 2012 (14:20)