[Noisebridge-discuss] Neural Net Workshop on 1/26/2010

[Mike Schachter]

Glad to help Jesse! There is no clear line between ANNs and BNNs,
and lots of Computational Neuroscientists and Computer Scientists
are keen on combining them.

The major benefit of ANNs is that there are backpropagation algorithms
guaranteed to effectively train them, whereas no such algorithms are
out there for BNNs. For training BNNs, some people look to biological
training routines that work like spike-timing-dependent-plasticity:

http://en.wikipedia.org/wiki/Spike-timing-dependent_plasticity

Recently some people came up with a "tempotron", a biologically-
plausible neuron that can be trained to recognize temporal sequences:

http://www.nature.com/neuro/journal/v9/n3/abs/nn1643.html

I'm sure over time the two fields will converge even further, looking
forward to talking more about it soon. Thanks for the questions!

  mike



On Mon, Jan 3, 2011 at 2:16 PM, Jesse welz <welzart at gmail.com> wrote:

> Mike,
>
> Thank you for the clarification. That totally makes sense and I kind
> of remember coming across some info explaining the differences of BNNs
> and ANNs. I'll bet the two types of networks could be combined
> although there is probably no use for such a combination.
>
> I will definitely try and make that workshop as I find A.I. absolutely
> fascinating. I've always wanted to sit in on one of the Machine
> Learning workshops but never had the chance. This gives me good excuse
> to make sure I show up.
>
> Gescykae
>
> On Mon, Jan 3, 2011 at 2:04 PM, Mike Schachter <mike at mindmech.com> wrote:
> > Hi Jesse,
> >
> > I don't have much familiarity with B.E.A.M robotics, but I did
> > read through this FAQ:
> >
> > http://faq.solarbotics.net/FAQ.html#q3
> >
> > The workshop is focusing on artificial neural networks, which
> > are not the same as biological neural networks. If B.E.A.M
> > robots are based on analog RC circuits, then they're more
> > biological NNs (BNNs) than artificial NNs (ANNs).
> >
> > There are many similarities between biological and artificial
> > neural networks that I'd be happy to talk about though. For
> > example the layered network structure is similar between the
> > two. The theory behind ANNs are more developed than for
> > BNNs, but the two fields play off eachother heavily.
> >
> > The difference between the two types of networks is in how
> > complex the model for an individual neuron is. A neuron in
> > an ANN is typically less complex than a neuron in a BNN,
> > which makes it easier to analyze mathematically and also
> > to simulate and train on a computer.
> >
> > I'd be happy to elaborate via the list or in person, the machine
> > learning group meets up every Wednesday at 7:30pm in the
> > back classroom.
> >
> >   mike
> >
> > On Mon, Jan 3, 2011 at 1:56 PM, Jesse welz <welzart at gmail.com> wrote:
> >>
> >> Hi Mike,
> >>
> >> Are the neural networks you're talking about here at all related to
> >> the neural networks that B.E.A.M. robotics are based on? I know
> >> nothing about neural networks other than what I've picked up tinkering
> >> with B.E.A.M. robotics. Even if they are not related I would love
> >> learn more and become Neural Network Aware. I am not a mathematician
> >> nor a programmer but have dabbled in both. Will I be able to follow
> >> along?
> >>
> >> Best,
> >> Gescykae
> >> _______________________________________________
> >> Noisebridge-discuss mailing list
> >> Noisebridge-discuss at lists.noisebridge.net
> >> https://www.noisebridge.net/mailman/listinfo/noisebridge-discuss
> >
> >
> _______________________________________________
> Noisebridge-discuss mailing list
> Noisebridge-discuss at lists.noisebridge.net
> https://www.noisebridge.net/mailman/listinfo/noisebridge-discuss
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.noisebridge.net/pipermail/noisebridge-discuss/attachments/20110103/da90a5ca/attachment-0003.html>