[ExI] ANN question
atymes at gmail.com
Wed Mar 15 19:26:18 UTC 2017
On Mar 15, 2017 11:00 AM, "William Flynn Wallace" <foozler83 at gmail.com>
On Wed, Mar 15, 2017 at 12:29 AM, Adrian Tymes <atymes at gmail.com> wrote:
> At any given instant, a neuron may either be firing or have a time
> until next firing - but a firing is a (mostly) discrete event, with
> one firing distinct from the next.
> If it was not distinct - if it was "always firing" - then there would
> be no such thing as "speed" of firing. Rather than toggle on then off
> at various rates, it would simply be always on.
There is a great deal of controversy about rate of firing, from about 1
every 6 seconds to 200 per second, up to a couple of thousand at the
fastest (for cortical neurons). By always firing I meant only that it
never rested more than a few seconds at the very most. Yes, discrete -
absolute refractory period.
Right. And during that refractory period, the output is 0, regardless of
how long it will be or has been to the next/previous firing.
So, again, that gives three states - not changing speed, slowing, increasing
No, two states: firing, or between firings.
You are declaring the acceleration to be a state, when the state is more
like the current position, and the simplification is that most ANNs don't
even consider velocity let alone acceleration.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the extropy-chat