[nengo-user] concept activation question

[Terry Stewart]

Hello Jessica,

Hmm, interesting.  I think the main question would be to sort out what
sort of algorithm you'd want that would do something like this.  Nengo
is great for taking an algorithm and implementing it in neurons, but
in order for that to work we need some sort of algorithm to start
with.

That said, the sort of thing you're describing seems like a natural
property of a lot of the vector-symbolic-architecture models that are
out there, and those can be very natural to implement in Nengo (that's
the core of what we call the Semantic Pointer Architecture).  For
example, take a look at what's been done with BEAGLE
<http://www.indiana.edu/~clcl/data.htm> for people just looking at the
algorithm (and not worrying about neural implementation).  For Nengo,
I think the closest we've done in that situation is some of Peter
Blouw's work, such as
<http://mindmodeling.org/cogsci2013/papers/0353/paper0353.pdf>.

Does that help?
Terry

On Fri, Mar 28, 2014 at 12:08 AM, Jessica Dinh <jd62 at zips.uakron.edu> wrote:
> Hello,
>
> I wanted to build a model that looks at the activation of a concept or
> behavior as an outcome of interactions among multiple processing systems,
> but I am having trouble building my model. I wanted to know if there are
> sample templates or helpful references that I can use to start my model.
>
> Ultimately, I wanted to create a model that simulates emergent concepts. For
> instance, the concept of 'cat' can emerge from the interaction of relevant
> information from visual (e.g., has fur, claws) and haptic (fuzzy) processing
> systems, but not irrelevant information from these systems (e.g., wags tail,
> drools, barks which should activate the concept of a dog instead of a cat).
> Also, whether a concept emerges would depend on how many pieces of
> information are available to the network. For example, 'cat' might emerge
> quickly when information from both visual and haptic systems are available,
> but slowly when limited information comes from visual and/or haptic systems.
>
> I am not quite sure where to start, in particular, the network configuration
> and how information should be presented for the network to learn these
> associations during initial training. Thank you very much in advance, and
> any help would be much appreciated.
>
> ~Jessica
>
>
> --
> Dr. Jessica E. Dinh
> Industrial/Organizational Psychology
> The University of Akron
> Arts & Sciences Building, Mailbox #20
> Akron, OH 44325
> jd62 at zips.uakron.edu
>
> _______________________________________________
> nengo-user mailing list
> nengo-user at ctnsrv.uwaterloo.ca
> http://ctnsrv.uwaterloo.ca/mailman/listinfo/nengo-user
>