JustNN neural networks
Stephen Wolstenholme
your data from txt, csv, xls, bmp or binary files with just a few
clicks. Grow a multi layer neural network that will learn from your
data. Validate while the network is learning. Then test or query the
network using new data to produce results and see which inputs are
really important. JustNN has no grid row, grid column limits, network
node or connection limits. It has no evaluation date or time limit
and it is entirely free.
--
Neural Planner Software Ltd www.NPSL1.com
EasyNN-plus. Neural Networks plus. www.easynn.com
SwingNN. Forecast with Neural Networks. www.swingnn.com
JustNN. Just Neural Networks. www.justnn.com
alvey
> Building neural networks from your data could not be simpler. Import
> your data from txt, csv, xls, bmp or binary files with just a few
> clicks. Grow a multi layer neural network that will learn from your
> data. Validate while the network is learning. Then test or query the
> network using new data to produce results and see which inputs are
> really important. JustNN has no grid row, grid column limits, network
> node or connection limits. It has no evaluation date or time limit
> and it is entirely free.
Has anyone used this? Recommendations?
alvey
--
Ignarus alio nomen secundum bestia.
Stephen Wolstenholme
<Yrrah-hel...@mail-elsewhere.ca> wrote:
>Stephen Wolstenholme <st...@tropheus.demon.co.uk>:
>
>> Building neural networks from your data could not be simpler.
>
>Has anyone here ever found a use for artificial neural networks?
>http://en.wikipedia.org/wiki/Artificial_neural_network
>
I have few thousand users of my commercial neural network products so
I must assume some people have found a use for them. The main uses
that I know are predicting stock, exchange rates and sports results.
There are some testimonials from users on the web sites.
I made JustNN free to help to increase the general awareness of the
use of artificial neural networks.
Steve
Stephen Wolstenholme
<Yrrah-hel...@mail-elsewhere.ca> wrote:
>Stephen Wolstenholme <st...@tropheus.demon.co.uk>:
>
>> The main uses
>> that I know are predicting stock, exchange rates and sports results.
>
>"stock, exchange rates and sports results" can be predicted?
>
It's never going to be 100% accurate but neural networks do a lot
better than just guessing. There is a horse racing example with both
JustNN and EasyNN-plus that gives 50% to 80% accurate results when
tested with real data.
A really good method is to set up a neural network that is trained
using the successful predictions of real experts. Getting the experts
to give you their data can prove to be the main hurdle.
Saxman
> A really good method is to set up a neural network that is trained
> using the successful predictions of real experts. Getting the experts
> to give you their data can prove to be the main hurdle.
It's not always the number of winners that count......but the starting
price (SP) that they are priced at.
Neural Networks are not very good at picking outsiders at good prices,
much in the same way that humans are not.
Stephen Wolstenholme
The great advantage of both humans and neural networks is that they do
a lot better than an "unskilled" selection. Neural networks can be
trained to included the starting price. The whole idea is to use a
software application to do the hard work. It's not unknown for a
neural network to pick an outsider but who would trust it?
Craig
> Building neural networks from your data could not be simpler...
Stephen;
I noticed JustNN is for MSOS. Is there anything comparable in the
linux/unix user space that you're aware of?
tia,
--
-Craig
Craig
>
> This link any use?
Thanks Gordon. I'll take a look later today. loll, huh? That's a new
resource for me. Much appreciated!
--
-Craig
Stephen Wolstenholme
wrote:
There are a few good applications for UNIX but they are often
complicated and very expensive. Most have been ported to Windows but
they are still complicated and very expensive.
The market for easy to use, low price and free neural network
applications is, so far as I know, limited to Windows.
So far as I am concerned the key is ease of use. You can get extensive
development systems and libraries for UNIX but they are not easy to
use.
Steve
Saxman
> The great advantage of both humans and neural networks is that they do
> a lot better than an "unskilled" selection. Neural networks can be
> trained to included the starting price. The whole idea is to use a
> software application to do the hard work. It's not unknown for a
> neural network to pick an outsider but who would trust it?
I only know of one user of Neural Networks (NN) to make a profit from
gambling and he used it successfully with the spread betting companies.
Horse rating data is far too noisy for a NN to handle IMO.
Most examples purveyed by NN software companies relate to ice cream
sales throughout the year and such silly stuff. The human brain can
work that out.
There's plenty of horseracing data out there to test.
I can send you a sample file if you like?
Stephen Wolstenholme
<john.h.willi...@btinternet.com> wrote:
>Stephen Wolstenholme wrote:
>
>> The great advantage of both humans and neural networks is that they do
>> a lot better than an "unskilled" selection. Neural networks can be
>> trained to included the starting price. The whole idea is to use a
>> software application to do the hard work. It's not unknown for a
>> neural network to pick an outsider but who would trust it?
>
>I only know of one user of Neural Networks (NN) to make a profit from
>gambling and he used it successfully with the spread betting companies.
>
>Horse rating data is far too noisy for a NN to handle IMO.
>
Neural networks work very well with noisy data. Training to a low
error takes longer with noisy data but the objective is to get an
advantage over just guessing. Even a small advantage can make a
profit.
>Most examples purveyed by NN software companies relate to ice cream
>sales throughout the year and such silly stuff. The human brain can
>work that out.
>
Another useful feature of good NN applications is that they can find
the relative importance of input measures. That could be useful in
sales. Even ice cream sales.
>There's plenty of horseracing data out there to test.
>
>I can send you a sample file if you like?
Thanks but I've got plenty of horse racing data.
chris
> I have few thousand users of my commercial neural network products so
> I must assume some people have found a use for them. The main uses
> that I know are predicting stock, exchange rates and sports results.
yeah, making a fortune with predicting stock exchange results is what every
idiot wants ;-).
Don't forget medical applications like analysing results and/or predicting
the change of having an illness.
The medical world uses NN more and more and couldn't do their job
without it. Almost every industry uses NN today.
The biggest problem is getting good test sets.
Thanks for making your software freeware.
My Artificial Intelligence ( BBS ) software is from the beginning.
That means DOS based and a very little W95.
You can still find it on
http://chris.mysteryfcm.co.uk/
Thanks again for making your software free available.
( and thanks to mysteryfm for still hosting this software )
Stephen Wolstenholme
>"Stephen Wolstenholme" wrote
>
>> I have few thousand users of my commercial neural network products so
>> I must assume some people have found a use for them. The main uses
>> that I know are predicting stock, exchange rates and sports results.
>
>yeah, making a fortune with predicting stock exchange results is what every
>idiot wants ;-).
In my experience some users predicting stock and currency prices have
dug very deeply into AI methods and then disappeared! They have
either been successful and want to keep their methods secret or they
have lost a lot of money and are embarrassed about it.
>Don't forget medical applications like analysing results and/or predicting
>the change of having an illness.
>The medical world uses NN more and more and couldn't do their job
>without it.
I have a lot of users of EasyNN-plus in the medical world.
>Almost every industry uses NN today.
>The biggest problem is getting good test sets.
Getting data for training, validating and testing can be a problem.
Many of my support calls are from users who expect too much from very
small data sets. Then there are the users who expect to be able to
predict next week lottery numbers because they have all the results
since the start of time.
>Thanks for making your software freeware.
>
>My Artificial Intelligence ( BBS ) software is from the beginning.
>That means DOS based and a very little W95.
My original stuff is pre-DOS. My very first neural network application
ran on a System 4 mainframe in the early 2970's. I'm showing my age :(
>You can still find it on
>http://chris.mysteryfcm.co.uk/
>
That's an interesting collection but a bit too much for me to explore.