Evolution of dendritic morphologies for pattern recognition in a passive neuron model

Mohammad Ziyad Kagdi, Rene Te Boekhorst

Research output: Chapter in Book/Report/Conference proceedingPublished conference contribution

Abstract

This research investigates the impact of dendritic morphology of a neuron on pattern recognition capacity and it addresses the question of how pattern recognition performance of a neuron can be determined by its dendritic structure and morphological properties. To systematically explore the dendritic space, a branching stochastic approach is used to create the dendritic structure so as to agree with the structure of a biologically realistic neuron. For every possible dendritic tree generated, there are 128 terminal nodes and 255 compartments which were further used to construct multicompartmental models. In each model, the number of terminal nodes and compartments is fixed so as to investigate parameters such as the compartment length, mean path length, axial resistivity, synaptic locations and the extent to which they affect pattern recognition capacity. The neurons are presented with an equal number of stored and novel binary input patterns where the ability of pattern recognition is determined by calculating the signal to noise ratio between amplitudes of Excitatory Post Synaptic Potentials (EPSPs). The objective of this approach is to analyse the types of input pattern in combination with morphological features which may strengthen or weaken the EPSP amplitudes. Finally, an evolutionary algorithm is used to explore the parameter space and to produce a variety of branching structures of dendrite which are good enough to recognise patterns. The study further identifies that the dendritic structure has significant impact on pattern recognition performance and that the neuronal performance is broadly affected by its synaptic locations.
Original languageEnglish
Title of host publicationComputational Intelligence Methods for Bioinformatics and Biostatistics
EditorsAndrea Bracciali, Giulio Caravagna
Place of PublicationUnited Kingdom
PublisherUniversity of Stirling
Pages64-69
Number of pages6
Publication statusPublished - 31 Aug 2016
EventCIBB 2016: Computational Intelligence Methods for Bioinformatics and Biostatistics - University of Stirling, Stirling, United Kingdom
Duration: 1 Sept 20163 Sept 2016
Conference number: 13
http://www.cs.stir.ac.uk/events/cibb2016/

Conference

ConferenceCIBB 2016
Country/TerritoryUnited Kingdom
CityStirling
Period1/09/163/09/16
Internet address

Keywords

  • Neuronal model
  • Passive properties
  • Dendritic morphology
  • Hebbian learning
  • Pattern recognition
  • Evolutionary algorithm

Fingerprint

Dive into the research topics of 'Evolution of dendritic morphologies for pattern recognition in a passive neuron model'. Together they form a unique fingerprint.

Cite this