The Proceedings of JSME annual Conference on Robotics and Mechatronics (Robomec)
Online ISSN : 2424-3124
2014
Session ID : 3P2-P05
Conference information
3P2-P05 A Computational Model for Local Processing Bias in Autism Spectrum Disorders(Neurorobotics & Cognitive Robotics)
Takakazu MORIWAKIYukie NAGAIMinoru ASADA
Author information
CONFERENCE PROCEEDINGS FREE ACCESS

Details
Abstract
It is known that people with Autism Spectrum Disorders (ASD) have a bias for local information processing, whereas normally developing people exhibit a global bias. This study examines its underlying mechanism inspired by an evidence from neuroscience studies. Our hypothesis is that an imbalance between excitation and inhibition neurons modifies a threshold for neural activities and thus changes a bias for local or global information. We employed a hierarchical neural network called neocognitron to examine the influence of the excitation/inhibition balance on the recognition of hierarchical compound letters (i.e., a global/large letter consisting of local/small letters). Our experiments demonstrated that hyper inhibitory connections made the network recognize a local letter better than a global letter, whereas a proper excitation/inhibition balance produced an opposite result. This empirically supports our hypothesis that the excitation/inhibition imbalance is a cause of a local processing bias in ASD.
Content from these authors
© 2014 The Japan Society of Mechanical Engineers
Previous article Next article
feedback
Top