Re: Brain protocol

[John Eikenberry [MSAI]]

On Wed, 19 Feb 1997, Juan J. Merelo wrote:

>       Anything else? Anything missing? What about some kind of
> feedback? Once the brain "decides", it should do something with the
> results of that decision. That usually happens in neural networks,
> where feedforward and training are separated. Besides, that feedback
> might be needed to increase fitness.
> 
>       What about:
> -(void) reThink: (id);    // Could be mull, or feedback.

I think this sounds like a good idea. I think this would definately make 
teaching the agents easier, not just for neural nets, but possibly for 
other types as well. Say with a reinforment learing method. The agent 
decides what it wants to do, does it, sees the results then considers how 
good it's decision was.

Of couse this could be done as part of the think method (consider how 
well it did the last time around), but I like the better control having
a separate method would allow.

John Eikenberry
address@hidden
______________________________________________________________
"A society that will trade a little liberty for a little order
 will deserve neither and lose both."
                                         --B. Franklin