BEGIN:VCALENDAR
VERSION:2.0
PRODID:talks.ox.ac.uk
BEGIN:VEVENT
SUMMARY:A simple neural network to structure thoughts using rapidly changi
 ng synapses - Sanjay Manohar (University of Oxford)
DTSTART;VALUE=DATE-TIME:20231130T143000Z
DTEND;VALUE=DATE-TIME:20231130T153000Z
UID:https://talks.ox.ac.uk/talks/id/3903e761-ef36-4264-a864-c74250e87ea5/
DESCRIPTION:A central organizing principle of human mental life is grammar
 . It allows us to separate the structure from the content of thought. Few 
 models explain how grammar may be implemented in neurons. We combined two 
 rapid Hebbian synaptic plasticity rules to demonstrate how neurons can imp
 lement simple grammar. The first rule associates neurons representing word
 s with neurons representing syntactic roles\, e.g. “dog” may associate
  with "subject" or "object". The second rule establishes the sequential or
 dering of roles (e.g. subject → verb → object)\, guided by predefined 
 syntactic knowledge. We find that\, like humans\, the network encodes and 
 retrieves grammatical sentences better than shuffled word-lists. It can se
 rialize a 'bag of words' to express an idea as a sentence. The network can
  model languages that rely on syntactic order\, but also order-free morphe
 mic languages. The model predicts the existence of syntactic and lexical p
 riming\, and can simulate evoked potentials recorded from EEG. When lesion
 ed\, the network exhibits classical symptoms of neurological aphasia\, inc
 luding dissociation between agrammatic and semantic aphasia\, unlike curre
 nt deep neural network language models. Crucially\, it achieves all this u
 sing an intuitive representation where words fill roles\, emulating struct
 ured cognition.\nSpeakers:\nSanjay Manohar (University of Oxford)
LOCATION:Sherrington Library\, off Parks Road OX1 3PT
TZID:Europe/London
URL:https://talks.ox.ac.uk/talks/id/3903e761-ef36-4264-a864-c74250e87ea5/
BEGIN:VALARM
ACTION:display
DESCRIPTION:Talk:A simple neural network to structure thoughts using rapid
 ly changing synapses - Sanjay Manohar (University of Oxford)
TRIGGER:-PT1H
END:VALARM
END:VEVENT
END:VCALENDAR