Modeling neural networks using process algebras

Colvin, Robert (2013). Modeling neural networks using process algebras. , Queensland Brain Institute, The University of Queensland.

Attached Files (Some files may be inaccessible until you login with your UQ eSpace credentials)
Name Description MIMEType Size Downloads
main.pdf Full text (open access) application/pdf 543.87KB 42
Author Colvin, Robert
Title Modeling neural networks using process algebras
School, Department or Centre Queensland Brain Institute
Institution The University of Queensland
Open Access Status Other
Publication date 2013-11-12
Start page 1
End page 31
Total pages 31
Language eng
Formatted abstract
Research involving artificial neural networks has tended to be driven towards efficient computation (e.g., in the domain of pattern recognition) or towards elucidating biological processes in the brain (by fitting simulations to experimental data). As our understanding of the biology of individual neurons and the brain as a whole has increased, so have neural network models become more complex, incorporating real-time behaviour, hybrid behaviour (discrete events alongside continuously changing variables), on top of complex system structure and dynamics. To date there has been relatively little effort in fully formally describing neural networks, with typical descriptions in the literature being a mixture of mathematical equations and natural language. This often hides or obscures important aspects of a particular model, and leaves a large conceptual gap between the model descriptions and the usually low-level programming code used to simulate them. In this paper we describe a formal notation and its behaviour which is suited for modelling neural networks. The language is a process algebra, which are well-established mathematical languages for describing highly concurrent (computer) systems and inter-process communication. The basic notions from process algebras are extended with local state variables for maintaining information about individual neurons and synapses, allowing both instantaneous and continuous changes to their values. The notation is used initially to formalise feedforward, backpropagation, and recurrent network behaviour, and then to formalise a more recent neural network model that includes real-time behaviour, differential equations, and complex spatial and temporal relationships between neurons.
Q-Index Status Provisional Code
Institutional Status UQ

Version Filter Type
Citation counts: Google Scholar Search Google Scholar
Created: Mon, 11 Nov 2013, 09:30:17 EST by Dr Robert Colvin on behalf of Queensland Brain Institute