In article <36F23B6B.814145E7 at axcess.co.uk>,
Karla Parussel <kparussel at axcess.co.uk> wrote:
>> First off, sorry if this is obvious, in the FAQ (is there one), an
> extremely ignorant question
> or just not relevant (comp.ai.neural-net couldnt really help). I am not
> actually trained in biology or neuroscience, my path has been more a :
>> Computer Science -> Artificial Life -> interest in Computational
>> Anyway. I have implemented a 'biologically plausible' neural tool kit.
> (I get bored in the evenings you see). These are basically spiky,
> integrate and fire, threshold neurons that leak their charge. I have
> been reading a lot of literature from a lot of places and this is the
> end result. Even though it now does pavolvian conditioning perfectly, I
> can't use it to implement competition (where a group of neurons will try
> to inhibit the other neurons while exciting themselves).
>> Anyway, what has started to become apparent is that I don't actually
> have any
> interneurons. Ooops. I know that they come in all shapes and sizes (like
> neurons generally
> do) and that the difference between interneurons (Golgi type II) and
> principle neurons (Golgi type I) is that the interneurons connect to
> outside the immediate area / cluster / whatever whereas principle
> neurons do. But is there any other difference? Someone on
> comp.ai.neural-nets said that the not all neurons intergrate, reach a
> threshold and fire. Would some just propagate any charge they received
> from the synapses on their dendrites immediately to their axon for
>> Would this be relevant to inhibitory interneurons that are said to be
> found quite a lot in competitive networks? I have looked all over the
> place, searched the net and read through my books but I just can't seem
> to find the answers. My maths isn't really that strong which probably
> limits me a lot when it comes to understanding large mathematical
> formuli. If anyone could point me in the right direction I would be
> really appreciative.
>> Anyway, sorry if this is completely off-topic etc etc I havn't really
> seen any other posts like this here but I am not sure where else to ask.
>> Thanks :)
For the latest on the properties of real neurons see the book:
Biophysics of Computation - Information Processing in Single Neurons by
Christof Koch; Oxford University Press - 1999
This is an excellent book by one of the leaders in computational neuroscience
The major conclusions are that neurons do not work by the "integrate and fire"
method and that some new approach to brain modeling is needed. The fact that
synapses are probabilistic working on average only 1/3 of the time means
that viewing action potential spikes as binary carriers of information is not
My own investigations relate to asynchronous multivalued logic as applied to
brain circuits. To learn more on this approach and for some information on
frog brains see my site at http://neurocomputing.org.
-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/ Search, Read, Discuss, or Start Your Own