Reconstruction, Identification and Implementation Methods for Spiking Neural Circuits

This work is motivated by the ongoing open question of how information in the outside world is represented and processed by the brain. Consequently, several novel methods are developed.

A new mathematical formulation is proposed for the encoding and decoding of analog signals using integrate-and-fire neuron models. Based on this formulation, a novel algorithm, significantly faster than the state-of-the-art method, is proposed for reconstructing the input of the neuron.

Two new identification methods are proposed for neural circuits comprising a filter in series with a spiking neuron model. These methods reduce the number of assumptions made by the state-of-the-art identification framework, allowing for a wider range of models of sensory processing circuits to be inferred directly from input-output observations. A third contribution is an algorithm that computes the spike time sequence generated by an integrate-and-fire neuron model in response to the output of alinear filter, given the input of the filter encoded with the same neuron model.

Nominated as an outstanding Ph.D. thesis by the University of Sheffield Reformulates, for the first time, the encoding performed by the integrate-and-fire neuron model as a problem of uniform sampling of an auxiliary function on a set of input independent time points Proposes two methodologies for identifying [Nonlinear Filter]-[Ideal IF] and [Linear Filter]-[Leaky IF] circuits Developes for the first time, a direct representation between the input and output of a linear filter, both encoded with an integrate-and-fire neuron model Includes supplementary material: sn.pub/extras Includes supplementary material: sn.pub/extras

Auteur
Dr Dorian Florescu is currently a Postdoctoral Research Fellow in the Department of Automatic Control and Systems Engineering at the University of Sheffield, working on the 'Digital Fruit Fly Brain' project, funded jointly by BBSRC and the National Science Foundation.

He was awarded the BEng degree in Systems Engineering from the Technical University of Iasi, Romania, in 2011 and the PhD degree in Automatic Control & Systems Engineering from the University of Sheffield

Contenu

Nomenclature.- Acronyms.- 1 Introduction.- 2 Time Encoding and Decoding in Bandlimited and Shift-Invariant Spaces.- 3 A Novel Framework for Reconstructing Bandlimited Signals Encoded by Integrate and-Fire Neurons.- 4 A Novel Reconstruction Framework in Shift-Invariant Spaces for Signals Encoded with Integrate-and-Fire Neurons.- 5 A New Approach to the Identification of Sensory Processing Circuits Based on Spiking Neuron Data.- 6 A New Method for Implementing Linear Filters in the Spike Domain.- 7 Conclusions and Future Work.- Bibliography.