On Jul 15, 10:03 pm, "Phil Sherrod" <PhilSher... at (no spam) REMOVEcomcast.net
wrote:
On 15-Jul-2009, Greg <he... at (no spam) alumni.brown.edu> wrote:

If I don't have enough a priori knowlefge to use parametric
regression, I have the comfort of being able to use NN
univeral approximators to implement arbitrary nonlinear
regression.

Agreed.  That is the strength of NN.

Starting in 1983 my original Star Wars radar classifier
dsesigns were based on combining readily interpretable
pairwise linear classifiers that were specifically designed
to separate predetermined clusters of different classes.
For the next 15 years I became more adept at constructing
techinques for fast learning and modification of classifiers
based on a huge, continually expanding measurement data
base. The constraints of fast learning, fast modification
and interpretability eventually lead to a change from MLP
to RBF architecture. However, for neither architecture was
the interpretability constraint amenable to standard
numerical minimization techniques.

Sounds like an interesting project.

Back to the A.I. discussion, I can't define "intelligence",
but I believe one characteristic of it
is the ability to infer (extrapolate) outcomes beyond the
direct experience of the observer.  If
that is true, then a NN is further from A.I. than a simple
polynomial function.

I don't agree with that comparison. My experiences with
polynomial extrapolation were much more disasterous than
those with NNs. The boundedness of sigmoids and Gaussians
make the difference.

I didn't mean blindly fitting a polynomial.  I meant fitting a polynomial if the actual function is
polynomial such as the trajectory of a ballistic object in a vacuum.  What I am trying to say is a
function that has a theoretical basis for fitting the data.

I've seen many cases where a MLP fits data well within the training range and then goes flying off
wildly if you extrapolate it outside the range.  All a MLP is is a technique for spline fitting: the
spline segments are formed by scaling, translating, and adding sections of the nonlinear transfer
function.  As with any spline fitted curve, it tends to go berserk beyond the end points.

--
Phil Sherrod
(PhilSherrod 'at' comcast.net)http://www.dtreg.com (Decision trees, Neural networks,  SVM)http://www.nlreg.com (Nonlinear Regression)

Recently someone (Steve?) mentioned looking into
extrapolating NNs. I forget the details.

Recently someone (Steve?) mentioned looking into
extrapolating NNs. I forget the details.

http://groups.google.com/group/comp.ai.neural-nets/msg/0fa51c8b34d890...

Hope this helps.

Recently someone (Steve?) mentioned looking into
extrapolating NNs. I forget the details.

http://groups.google.com/group/comp.ai.neural-nets/msg/0fa51c8b34d890...

Hope this helps.

Recently someone (Steve?) mentioned looking into
extrapolating NNs. I forget the details.

http://groups.google.com/group/comp.ai.neural-nets/msg/0fa51c8b34d890...

Hope this helps.

Can someone please explain? I have found the original thread here -

http://groups.google.com/group/comp.ai.neural-nets/browse_thread/thre...

It looks like two people - Phil Sherrod, and Greg Heath, are talking
with someone named Steve, but I can't see any of his posts, the thread
contains only post of Phil and Greg, but where are the original posts
of Steve? I can find only quotations of his words but where are his
posts? is he a ghost?

On Sat, 18 Jul 2009 02:44:42 -0700 (PDT), Greg





he... at (no spam) alumni.brown.edu> wrote:
On Jul 17, 5:08 am, Sigmoid 2009 <sigmoid2... at (no spam) gmail.com> wrote:
Recently someone (Steve?) mentioned looking into
extrapolating NNs. I forget the details.

http://groups.google.com/group/comp.ai.neural-nets/msg/0fa51c8b34d890....

Hope this helps.

Can someone please explain? I have found the original thread here -

http://groups.google.com/group/comp.ai.neural-nets/browse_thread/thre....

It looks like two people - Phil Sherrod, and Greg Heath, are talking
with someone named Steve, but I can't see any of his posts, the thread
contains only post of Phil and Greg, but where are the original posts
of Steve? I can find only quotations of his words but where are his
posts? is he a ghost?

Unfortunately, Steve put a time limit on his posts.
After a specified number of days his posts were
automatically deleted.

Greg

No responses from me as I've been ill but I'm back now.

There is no time limit on my posts. I don't use a service with such
facilities. So far as I know Google just copies Usenet. I've no idea
what has happened to my posts.

On Jul 17, 5:08 am, Sigmoid 2009 <sigmoid2... at (no spam) gmail.com> wrote:
Recently someone (Steve?) mentioned looking into
extrapolating NNs. I forget the details.

http://groups.google.com/group/comp.ai.neural-nets/msg/0fa51c8b34d890...

Hope this helps.

Can someone please explain? I have found the original thread here -

http://groups.google.com/group/comp.ai.neural-nets/browse_thread/thre...

It looks like two people - Phil Sherrod, and Greg Heath, are talking
with someone named Steve, but I can't see any of his posts, the thread
contains only post of Phil and Greg, but where are the original posts
of Steve? I can find only quotations of his words but where are his
posts? is he a ghost?

Unfortunately, Steve put a time limit on his posts.
After a specified number of days his posts were
automatically deleted.

Greg

On Jul 18, 6:56 am, Stephen Wolstenholme <st... at (no spam) tropheus.demon.co.uk
wrote:
On Sat, 18 Jul 2009 02:44:42 -0700 (PDT), Greg





he... at (no spam) alumni.brown.edu> wrote:
On Jul 17, 5:08 am, Sigmoid 2009 <sigmoid2... at (no spam) gmail.com> wrote:
Recently someone (Steve?) mentioned looking into
extrapolating NNs. I forget the details.

http://groups.google.com/group/comp.ai.neural-nets/msg/0fa51c8b34d890...

Hope this helps.

Can someone please explain? I have found the original thread here -

http://groups.google.com/group/comp.ai.neural-nets/browse_thread/thre...

It looks like two people - Phil Sherrod, and Greg Heath, are talking
with someone named Steve, but I can't see any of his posts, the thread
contains only post of Phil and Greg, but where are the original posts
of Steve? I can find only quotations of his words but where are his
posts? is he a ghost?

Unfortunately, Steve put a time limit on his posts.
After a specified number of days his posts were
automatically deleted.

Greg

No responses from me as I've been ill but I'm back now.

There is no time limit on my posts. I don't use a service with such
facilities. So far as I know Google just copies Usenet. I've no idea
what has happened to my posts.

Sorry if I am wrong but I'm pretty sure that
some of your posts had Google time limits
on them.

On Sat, 18 Jul 2009 04:44:06 -0700 (PDT), Greg





he... at (no spam) alumni.brown.edu> wrote:
On Jul 18, 6:56 am, Stephen Wolstenholme <st... at (no spam) tropheus.demon.co.uk
wrote:
On Sat, 18 Jul 2009 02:44:42 -0700 (PDT), Greg

he... at (no spam) alumni.brown.edu> wrote:
On Jul 17, 5:08 am, Sigmoid 2009 <sigmoid2... at (no spam) gmail.com> wrote:
Recently someone (Steve?) mentioned looking into
extrapolating NNs. I forget the details.

http://groups.google.com/group/comp.ai.neural-nets/msg/0fa51c8b34d890...

Hope this helps.

Can someone please explain? I have found the original thread here -

http://groups.google.com/group/comp.ai.neural-nets/browse_thread/thre...

It looks like two people - Phil Sherrod, and Greg Heath, are talking
with someone named Steve, but I can't see any of his posts, the thread
contains only post of Phil and Greg, but where are the original posts
of Steve? I can find only quotations of his words but where are his
posts? is he a ghost?

Unfortunately, Steve put a time limit on his posts.
After a specified number of days his posts were
automatically deleted.

Greg

No responses from me as I've been ill but I'm back now.

There is no time limit on my posts. I don't use a service with such
facilities. So far as I know Google just copies Usenet. I've no idea
what has happened to my posts.

Sorry if I am wrong but I'm pretty sure that
some of your posts had Google time limits
on them.

I don't know how to set a time limit. It may have been a default
setting but I can't remember ever visiting Google to make any
settings.

Where are the settings I need to check/change?

Is neural network same as artificial intelligence (AI) ?

If my application is base on a neural network for recognizing things,
for example objects in a picture, can I tell people that my
application is base on artificial intelligence (AI) ?or is it
something else?

Usually when I read in the news about a system that is based on an AI,
what does it means? does it means that it's based on a neural networs?
or does it means something else?

How can I keep one of my messages from being archived on Google
Groups?

To prevent a message from displaying after the usual seven days, or
from being searchable on Google Groups after being removed, make sure
it contains the text
'X-No-Archive:' either in the message header or as the very first text
in the first line
of the message body.