[nengo-user] Questions about nengo using neuronevolution technique

[Terry Stewart]

Hello Hugh,

We've been looking at a few different approaches for adjusting
topology over time, but at the moment we've really only been doing
things with connection weights.  That said, one of the nice things
about Nengo and the NEF is that you can use it to define and develop
an initial topology and connection weights, and then once you have
that network you can use it as the starting point for any other
learning algorithm, such as the neuroevolution methods you mention.
We find that can be a big improvement over just starting with random
topology/connection weights.

Terry

On Wed, Aug 6, 2014 at 3:05 PM, 吴乃龙 <478345869 at 163.com> wrote:
> Hi,
>      I am a phd candidate in robotics and AI. I find that nengo helps a lot
> in research. In human brain ,there are so many neurons, but only part of
> them are active. When the enviroment or the motion change, the neural
> network will alter too. According to neuroevolution methods(NEAT), the
> neural network can simultaneoulsy evolves both network structure (topology)
> and connection weights(NEAT). Does nengo support to build NEAT based
> network?  If so , could you give me some advice on how to build this kind of
> network? I will be very glad to get your advice.Thank you.
> Hugh
>
>
>
> _______________________________________________
> nengo-user mailing list
> nengo-user at ctnsrv.uwaterloo.ca
> http://ctnsrv.uwaterloo.ca/mailman/listinfo/nengo-user
>