[nengo-user] Resetting a Network

[Daniel Rasmussen]

Hi Brian,

It is possible to reset the simulator without rebuilding the network, by
calling network.reset().  That will reset all the neural states to random
initial conditions.  However, that kind of resetting will not give you a
lot of trial to trial variability, because most networks are not very
sensitive to those initial conditions (that's one of the strengths of the
NEF).  If you really want to explore the range of performance of a model,
then you do need to rebuild the whole thing, as you are doing now.  It's
during that build process that the encoders, decoders, and neuron
parameters are generated (that's why it's a bit slower, as you note), which
is where most of the variability comes from in a model.

Daniel


On 13 June 2014 11:15, Brian Krainer <bkrainer731 at gmail.com> wrote:

> Hello,
>
> Is there a way to 'reset' a network without having to rebuild it? I'm
> running an experiment and each trial I just reload the network which is a
> little bit slow. If I add a seed then each trial produces the same exact
> output which isn't want to be happening. After a network has been built is
> it deterministic?
>
> Thanks,
> Brian
>
> _______________________________________________
> nengo-user mailing list
> nengo-user at ctnsrv.uwaterloo.ca
> http://ctnsrv.uwaterloo.ca/mailman/listinfo/nengo-user
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://artsservices.uwaterloo.ca/pipermail/nengo-user/attachments/20140613/aa3aeab7/attachment-0002.html>