Grammatical structure in neural activation: Bridging discrete and continuous computation in the mind/brain

old_uid8925
titleGrammatical structure in neural activation: Bridging discrete and continuous computation in the mind/brain
start_date2010/06/16
schedule17h
onlineno
location_infosalle des actes
detailse Département d'Etudes Cognitives de l'ENS vous invite à la conférence de clôture de la Chaire Blaise Pascal
summaryA central element of the mind-brain problem can be stated, ‘how can the continuous, numerical nature of neural computation give rise to the discrete, categorical and abstract nature of mental computation?’ A homomorphism between vector space operators and symbolic functions provides the foundation of an approach to resolving this dilemma. The superpositional character of neural representations (activation patterns or vectors) helps explain why mental sub-processes need to compute discrete outputs. The approach is illustrated by computational models of neural networks that contain grammatical knowledge and discretization processes, and compute grammatical representations. These models offer explanations for a number of general empirical findings concerning speakers’ production of phonological sequences.
responsibles<not specified>