Associative spatial networks: Artificial cognition of space using neural networks with spectral graph theory

Harding, John

Abstract

This work stems from a backdrop of cybernetics and associative computing closely related to the artificial simulation of the human brain. Using associative techniques, a system to comprehend complex inputs in order to make appropriate architectural interventions is proposed.

A combination of abstract representations of space (graph theory) and a time based neural network clustering model (growing neural gas) are combined to create new spatial configurations based on pattern recognition and relationships formed from differences.

• Thesis Type: Dissertation
• Keywords: self-organising maps, graph spectra, growing neural gas, spatial cognition, feedback
• Public URL: https://uwe-repository.worktribe.com/output/1018920
• Award Date: Jan 1, 2008