Connectionists: Short-term plasticity to test hypotheses

[Andrea Soltoggio]

Dear All,


I’d like to point your attention to a new interpretation of the role of
short-term plasticity. In a study about to appear in the journal Biological
Cybernetics, and titled, “*Short-term plasticity as cause-effect hypothesis
testing in distal reward learning*” --preprint available at
http://arxiv.org/abs/1402.0710 -- I suggest a model in which transient
weight changes serve to test hypotheses when learning with distal rewards
(or also credit assignment problem).


Only tested hypotheses (i.e. verified correct over many trials) are
converted in long-term plastic changes, demonstrating how true cause-effect
relationships can be discovered without affecting the existing weight
structures, even with highly ambiguous information flow due to distal
rewards and overlapping stimuli and actions.


I hope you find it an interesting read. I would highly appreciate any
comment, discussion, criticism or alternative idea. I’m interested in
particular in discovering more biological studies supporting this new
interpretation.


Best regards,


Andrea Soltoggio


-- 
Dr.  Andrea Soltoggio
Haslegrave Building, N.2.03
Loughborough University
LE11 3TU
UK

Phone: +44 (0) 1509 635748
Email: a.soltoggio at lboro.ac.uk
Web:
http://www.lboro.ac.uk/departments/compsci/staff/dr-andrea-soltoggio.html
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/connectionists/attachments/20140813/82a9d45e/attachment.html>