Abstract
Background/Introduction: In this paper, the neural network class ofSelf-Organising Maps (SOMs) is investigated in terms of its theoretical andapplied validity for cognitive modelling, particularly of neurodevelopmentaldisorders. Methods: A modified SOM network type, with increased biological plausibility,incorporating a type of cortical columnar oscillation in the form of anoscillating Topological Neighbourhood (TN), is introduced and applied alongsidethe standard SOM. Aspects of two neurodevelopmental disorders, autism andschizophrenia, are modelled using SOM networks, based on existingneurocomputational theories. Both standard and oscillating-TN SOM training isemployed with targeted modifications in the TN width function. Computersimulations are conducted using revised versions of a previously introducedmodel (IPSOM) based on a new modelling hypothesis. Results/Conclusions: The results demonstrate that there is strong similaritybetween standard and oscillating-TN SOM modelling in terms of map formationbehaviour, output and structure, while the oscillating version offers a morerealistic computational analogue of brain function. Neuroscientific andcomputational arguments are presented to validate the proposed SOM modificationwithin a cognitive modelling framework.