A SOM-based validation approach to a neural circuit theory of autism

Abstract

The neural network class of self-organizing maps (SOMs) is a promising cognitive modeling tool in the study of the autistic spectrum pervasive developmental disorder. This work offers a novel validation of Gustafsson's neural circuit theory, according to which autism relates to formation characteristics of cortical brain maps. A previously constructed spatial SOM behavioral model is used here as a cognitive model, and by incorporating formation deficiencies related to the topological neighborhood (TN) function. The resulting cognitive SOM maps, being sensitive to the width of TN during SOM formation, point to a model that exhibits marked behavioral characteristics of autism. The simulation results support the causal hypothesis that associates autistic behavior with certain functional and structural characteristics of the human nervous system and, specifically, Gustafsson's theoretical proposition of the role of inhibitory lateral feedback synaptic connection strengths in autism. © 2012 Springer-Verlag.