On Sat, 11 Jul 2009 20:28:16 -0000,  Gunnar wrote:

>>
>Hi Steve
>1. Can you add a feature that expands the scale around the validation error so
that the validation minimum value can be more accurately determined?

Which scale?

>2. Often validation errors have more than one minimum. Can priority #4 be set
for say the third validating error increase?

The validating error is the average error for all validating example
rows. Individual minimums are too difficult to process in the current
development. I will put your request on the list for future
developments.

Steve

A bug in EasyNN-plus has been found that causes a crash when using the
Edit menu. The cause of the problem has been found (thanks Peter) but
it has not been fixed yet. It is top of the bug pile and will be fixed
soon. For now avoid right clicking row names.

Steve

--
Neural Planner Software Ltd                  www.NPSL1.com

EasyNN-plus V11.0d beta 1 is available for testing.

Some parameter validation macros have been changed. They are used in
hundreds of different places so a beta test is needed.

Other changes are as follows:-

1. Added Edit > Example Rows > Sort Example Rows

2. Added Edit > Example Rows > Sort Example Rows by Name

3. Added sortable format checkbox to the Insert Date & Time dialog.

4. Prioritize the validating stops.

5. Changed Insert Date & Time so that it works much faster in text
mode columns.

6. Changed image decoding to improve near miss resolution and speed.

7. Fixed a problem that could cause a crash after right clicking the
row name.

8. Fixed a problem that caused a validating cycle to run before any
training cycles.

9. Fixed a problem with F3 default action searching for '?'.

10. Fixed a problem caused by F2 and F3 interacting with Edit > Goto
Next Used.

11. Fixed a problem with Insert > * when no row or column has been
selected.

The download location is http://www.easynn.com/beta1104/ennplus.exe

For registered users it will work with your current registration code.
Unregistered users can get a temporary registration code by email from
support@...

Steve

--
Neural Planner Software Ltd                  www.NPSL1.com

EasyNN-plus V11.0d beta 2 is available for testing.

All the beta 1 bugs reported have been fixed.

The download location is http://www.easynn.com/beta1104/ennplus.exe

For registered users it will work with your current registration code.
Unregistered users can get a temporary registration code by email from
support@...

Steve

On Mon, 20 Jul 2009 17:51:44 +0100,  I wrote:

>A bug in EasyNN-plus has been found that causes a crash when using the
>Edit menu. The cause of the problem has been found (thanks Peter) but
>it has not been fixed yet. It is top of the bug pile and will be fixed
>soon. For now avoid right clicking row names.
>
>Steve

The fix for this problem, other bugs and some enhancements is
available in the latest beta at
http://www.easynn.com/beta1104/ennplus.exe

Steve

There is a problem with cutting text or image columns that can result
in the text or image coding being lost. If the cut column is then
pasted the text or image will be missing. This problem is caused by
deleting the column.

The problem will be resolved as soon as possible. For now with text or
image columns use Copy & Paste rather than Cut & Paste and exclude the
column rather than deleting it.

Steve

EasyNN-plus V11.0e is available from today.

The changes are as follows:-

1. Edit > Example Rows > Sort Example Rows by Type has been added.

2. Edit > Example Rows > Sort Example Rows by Value has been added.

3. Edit > Example Rows > Sort Example Rows has been removed.

4.  Edit > Cut has been changed so that it will not delete Image or
text mode columns unless the selected column is set to type Exclude.

5. Delete has been changed so that it always prompts if a Text or
Image mode column is selected even if the column is set to type
Exclude.

The full version has already been uploaded. Email support@...
for the download location.

The trial version will be available later today.

Steve

I just found your products and am not certain how to proceed with which product
to solve my issue.  I am working in visual basic.
I have three clusters of processed arrays of data (each array cluster is for a
specific condition of a processing system).  The number of arrays in each
cluster will not change.  I need to step through each array in each cluster one
element at a time, feeding the inputs of a neural network, then storing the
results of each cluster in its own array.  I need to recall the results of the
neural network outputs independent of the processing at a later time (when a GUI
is opened, etc.).  If I could reference each element in the array via timestamp,
that would be nice, but not necessary.
    TIA,
      DAS

On Tue, 25 Aug 2009 06:33:35 -0000,  you wrote:

>I just found your products and am not certain how to proceed with which product
to solve my issue.

Have a look at the script facilities in EasyNN-plus. Scripts and
macros are described in the Help file.

Steve

I've got my EasyNN neural network working and producing results almost as good
as a linear regression system I already had. I'd like now to be able to start
tweaking with my network to try different things out with the aim of improving
it. I don't know a lot about neural networks though. Can anyone recommend a book
or online resource which is compatible with the things EasyNN offers? The manual
is OK but tells you how to use the software.

There were a number of recommendations given (some by me included) in
the Forum, sometime back, after asking the same question.

P. Dickens


grimrobuk wrote:

> I've got my EasyNN neural network working and producing results almost
> as good as a linear regression system I already had. I'd like now to
> be able to start tweaking with my network to try different things out
> with the aim of improving it. I don't know a lot about neural networks
> though. Can anyone recommend a book or online resource which is
> compatible with the things EasyNN offers? The manual is OK but tells
> you how to use the software.
>
>
>


[Non-text portions of this message have been removed]

I am working with visual basic, is there a quick and easy way to do the
following with your NN family?
I am processing my arrays of data with my formulas and storing the results into
new arrays in memory (not messing with CSV files at all), now I need your
program to sequence through each array to determine weights, etc.
It does not have to be a continuously running task, but if it could be, then
that would be even better.
     Thanks,
          DAS

Here is a part of the content of one letter I wrote in reply to another
question, but regarding books, in the Forum, which I thought you might
miss because it was at the end of the letter.

................Your best scope is to use NN reading resources and be
prepared to go back twenty years plus for books and material, there is
no such thing as an out of date
book in Neural Networks, this is a relatively modern science and 95% of
what has been written on the subject, has been written in the last 30
years.  Some of the
best stuff was written between 1985 and 1995 during the time when
Computers were taking off (286, 386, and 486) and Software was being
written for them,
and frankly they still haven't been bettered; even though NN's have
moved on and evolved since then.

You would also be better off looking through the bibliographies of NN
books you have access to such as in a Library.  The IEEE also has a good
bibliographic list.  A good reference library will also give you a good
starting list.

Here is the URL for the IEEE publication shop where you can buy books
and  journals on NN, it is highly specialised being IEEE, but it will
give you an idea of
the range and scope in just one specific field (Electrical and
Electronics).   There are other similar NN activities in other sciences
and professions such as Finance,
Trading and Medicine.

http://shop.ieee.org/ieeestore/


I would also add that some NN books are extremely technical, both from a
NN design point of view and from a statistical and numeracy point of
view.  Try to find a listing that  has a categorisation system such as
beginner, intermediate and advanced.  Some books have two or even three
sections, covering each category.  Other books have been written using a
specific set of NN software, and that can make it very awkward to
follow, because most NN software have different interfaces, but will
provide you with some assistance.  One such book that I can thoroughly
recommend is:-

Introduction to Neural Networks  -  Design, Theory and Applications.
by Jeannette Lawerence.
ISBN 1-883157-00-5

It is a seminal piece of work, that to my knowledge has gone through 6
editions and 12 printing editions, that tells it's own story.  It does
start right at the beginning with more advanced sections to be used as
your experience grows.  J. Lawrence is a Physicist from the CALTECH
stable, who developed NN's, so her knowledge is born out of hard graft
at the cutting edge, and her writing style exemplifies that, knowing how
difficult it can be, so as a beginner you should be very comfortable
with her book, and should remain your reference source for many years.

Hope that helps.

P. Dickens


grimrobuk wrote:

> I've got my EasyNN neural network working and producing results almost
> as good as a linear regression system I already had. I'd like now to
> be able to start tweaking with my network to try different things out
> with the aim of improving it. I don't know a lot about neural networks
> though. Can anyone recommend a book or online resource which is
> compatible with the things EasyNN offers? The manual is OK but tells
> you how to use the software.
>
>
>


[Non-text portions of this message have been removed]

On Wed, 26 Aug 2009 14:00:40 -0000,  you wrote:

>I am working with visual basic, is there a quick and easy way to do the
following with your NN family?
>I am processing my arrays of data with my formulas and storing the results into
new arrays in memory (not messing with CSV files at all), now I need your
program to sequence through each array to determine weights, etc.
>It does not have to be a continuously running task, but if it could be, then
that would be even better.
>    Thanks,
>         DAS
>

There are no facilities that will access arrays in memory and I
wouldn't really know where to start producing such a facility. I
suppose it would require some sort of sharing to be established. This
may be possible but I don't know how. Even if I could work out how
access shared memory objects I would have lots of problems because I
don't know anything about Visual Basic and EasyNN-plus doesn't use
arrays for data storage. Even applications that do share data use
external objects stored in files. When I think a bit deeper access to
objects in memory would have all sorts of low level security hurdles
as well.

Steve

I see what you are saying - - I will move it to VB dotNet and see what the
minimum objects and references are required to do the task...
     Thanks,
          DAS


--- In easynn@yahoogroups.com, Stephen Wolstenholme <steve@...> wrote:
>
> On Wed, 26 Aug 2009 14:00:40 -0000,  you wrote:
>
> >I am working with visual basic, is there a quick and easy way to do the
following with your NN family?
> >I am processing my arrays of data with my formulas and storing the results
into new arrays in memory (not messing with CSV files at all), now I need your
program to sequence through each array to determine weights, etc.
> >It does not have to be a continuously running task, but if it could be, then
that would be even better.
> >    Thanks,
> >         DAS
> >
>
> There are no facilities that will access arrays in memory and I
> wouldn't really know where to start producing such a facility. I
> suppose it would require some sort of sharing to be established. This
> may be possible but I don't know how. Even if I could work out how
> access shared memory objects I would have lots of problems because I
> don't know anything about Visual Basic and EasyNN-plus doesn't use
> arrays for data storage. Even applications that do share data use
> external objects stored in files. When I think a bit deeper access to
> objects in memory would have all sorts of low level security hurdles
> as well.
>
> Steve
>

EasyNN-plus V12.0a beta 3 is available for testing.

The changes in V12.0a are as follows:-

1. The main change in this version are to take advantage of multi
core/processor systems. Learning runs in one to four processor.

2. Stop on decrease in validating score added to Controls dialog.

3. Edit > Example Rows > Maximize Validating Rows has been added.

4. All the stops in the control dialog have been moved to the same
group.

5. Learning thread start verification has been improved.

The download location is http://www.easynn.com/beta1201/ennplus.exe

The beta will work with your current registration code.

Unregistered users can get a temporary registration code by email from
support@...

Steve

EasyNN-plus 12.0a Beta 4 is now available.

The changes are as follows:-

1.  AutoSave has been changed so that file failures do not stop
learning. File failures only stop saving to the file that has failed.
Other files continue to be saved.

2.  Learning thread control has been moved to the Controls dialog so
that 1 to 8 learning threads can be started.

3. The problem with clipping columns has been resolved so the
restriction of the maximum number of columns has been removed.

The beta 4 setup file can be downloaded from
http://www.easynn.com/beta1201/ennplus.exe

If you are a registered user beta 4 will work with your current
registration code.

Unregistered users can get a temporary registration code by email from
support@...

Hi Steve
I have an intel i7 running xp 64 bit software. I tried to run 2 threads and
got junk. What am I missing?

On Mon, Oct 12, 2009 at 4:01 AM, Stephen Wolstenholme <
steve@...> wrote:

>
>
> EasyNN-plus 12.0a Beta 4 is now available.
>
> The changes are as follows:-
>
> 1. AutoSave has been changed so that file failures do not stop
> learning. File failures only stop saving to the file that has failed.
> Other files continue to be saved.
>
> 2. Learning thread control has been moved to the Controls dialog so
> that 1 to 8 learning threads can be started.
>
> 3. The problem with clipping columns has been resolved so the
> restriction of the maximum number of columns has been removed.
>
> The beta 4 setup file can be downloaded from
> http://www.easynn.com/beta1201/ennplus.exe
>
> If you are a registered user beta 4 will work with your current
> registration code.
>
> Unregistered users can get a temporary registration code by email from
> support@... <support%40easynn.com>
>
>
>



--
Regards
Gunnar Hurtig, III
(760)-469-2941
(408)-838-5813 Cell
www.incites.com


[Non-text portions of this message have been removed]

On Mon, 12 Oct 2009 14:01:32 -0700,  you wrote:

>Hi Steve
>I have an intel i7 running xp 64 bit software. I tried to run 2 threads and
>got junk. What am I missing?
>
I've no idea.

Can you tell me anything about the junk?

If you have anything that could help me diagnose the problem please
send it.

Please don't send any attachments to this group as it will filter them
out. You can email directly to steve@... if you need to attach
anything?

Steve

Anyone out there using Intel I7 965? I have two and experience inconsistencies
between the two. Same processor (I7-965) same OS (xp64bit) same settings same
ver 11-e. When plotting the Graph gp. I get a Validation error with spikes on
computer 1 and a smooth curve on Computer 2.
Any undocumented settings to be aware of?

On Sun, 18 Oct 2009 07:58:21 -0000,  you wrote:

>Anyone out there using Intel I7 965? I have two and experience inconsistencies
between the two. Same processor (I7-965) same OS (xp64bit) same settings same
ver 11-e. When plotting the Graph gp. I get a Validation error with spikes on
computer 1 and a smooth curve on Computer 2.
>Any undocumented settings to be aware of?
>

There must be some difference in the two *.tvq files at the start of
learning and validating.

Steve

--- On Fri, 10/16/09, Abdullah Farouk <abdullahfarouk@...> wrote:


From: Abdullah Farouk <abdullahfarouk@...>
Subject: [Career_egypt] نأسف لهذه الصورة والتي أبكت
حجر
To:
Date: Friday, October 16, 2009, 5:38 AM


 






نأسف لهذه الصورة


























 
 
نسعى لتوزيع أكثر من
 مليون رسالة إليكترونية
لإحياء عملية المقاطعة من جديد !!!
ليكن شعارنا
( من ترك شيئاً لله عوّضه الله خيراً منه )
 







 

صورة تبكي الحجر !!

صورة من صور مآسينا وما أكثرها!
صورة مؤلمة تثير الشجون وتسيل الدموع
كــــــان على ما يبدو يمشي مسرورا على
دراجته ببراءة الطفولة...
ولكن أيادي الغدر لم ترحـــم طفولته ....

لا حول و لا قوة إلا بالله!!

اللهم هون عليهم، و كن معهم و لا تكن عليهم!
اللهم ألطف بنا و بهم!
اللهم لا تؤاخذنا على غفلتنا و قسوة قلوبنا!
يا أكرم الأكرمين!

حسبي الله ونعم الوكيل
لا تنسوهم من دعائكم
 
 
 
لا تنسوا الهدف وهو توزيع أكثر من
 مليون رسالة إلكترونية
لإحياء عملية المقاطعة من جديد !!!
خذ بها و أرسلها وانشرها لكل من تعرف
 






 


















[Non-text portions of this message have been removed]

We can't read you message.

Please note that this email group is set to text only.

Steve

I know an Egyptologist who can read it, he specialises in Pharaohs'
Tombs.

Peter

Stephen Wolstenholme wrote:

> We can't read you message.
>
> Please note that this email group is set to text only.
>
> Steve
>
>
>


[Non-text portions of this message have been removed]

I was wondering what goes on during a query from file.  I have a net with about
200 inputs and about 50 nodes.  A query from file of about 400 input rows takes
around 5 hours.  I noticed the actual querying only takes about 10 minutes but
there is some kind of preprocessing that takes 5 hours.  I checked the process
and it is well within my memory capacity and running the processor full time. 
What is the preprocessing doing?

On Tue, 20 Oct 2009 05:02:02 -0000,  you wrote:

>I was wondering what goes on during a query from file.  I have a net with about
200 inputs and about 50 nodes.  A query from file of about 400 input rows takes
around 5 hours.  I noticed the actual querying only takes about 10 minutes but
there is some kind of preprocessing that takes 5 hours.  I checked the process
and it is well within my memory capacity and running the processor full time. 
What is the preprocessing doing?
>

That is much too slow. If you can send a copy of the *.tvq file and
the query file I will find out what the problem is. Please send the
files as attachments to support@...

Steve

Sorry I forgot to mention I have about 50,000 training rows which is probably a
major part of the slowdown.  The tvq file is 250 mb.

--- In easynn@yahoogroups.com, Stephen Wolstenholme <steve@...> wrote:
>
> On Tue, 20 Oct 2009 05:02:02 -0000,  you wrote:
>
> >I was wondering what goes on during a query from file.  I have a net with
about 200 inputs and about 50 nodes.  A query from file of about 400 input rows
takes around 5 hours.  I noticed the actual querying only takes about 10 minutes
but there is some kind of preprocessing that takes 5 hours.  I checked the
process and it is well within my memory capacity and running the processor full
time.  What is the preprocessing doing?
> >
>
> That is much too slow. If you can send a copy of the *.tvq file and
> the query file I will find out what the problem is. Please send the
> files as attachments to support@...
>
> Steve
>

If the network has already been trained the number of training rows
should not slow down querying. A query is only using one querying row
at a time. Try adding a querying row at the bottom of the grid and see
how long it takes to do a manual query with that row.

Steve

On Fri, 23 Oct 2009 01:49:14 -0000,  you wrote:

>Sorry I forgot to mention I have about 50,000 training rows which is probably a
major part of the slowdown.  The tvq file is 250 mb.
>
>--- In easynn@yahoogroups.com, Stephen Wolstenholme <steve@...> wrote:
>>
>> On Tue, 20 Oct 2009 05:02:02 -0000,  you wrote:
>>
>> >I was wondering what goes on during a query from file.  I have a net with
about 200 inputs and about 50 nodes.  A query from file of about 400 input rows
takes around 5 hours.  I noticed the actual querying only takes about 10 minutes
but there is some kind of preprocessing that takes 5 hours.  I checked the
process and it is well within my memory capacity and running the processor full
time.  What is the preprocessing doing?
>> >
>>
>> That is much too slow. If you can send a copy of the *.tvq file and
>> the query file I will find out what the problem is. Please send the
>> files as attachments to support@...
>>
>> Steve
>>
>
>
>
>
>------------------------------------
>
>Yahoo! Groups Links
>
>
>

There is a minor update to EasyNN-plus available today.

The changes are as follows:

1. A problem with spurious new line characters in LSO scripts has
been fixed.

2. LSO has been simplified by setting fixed cycles before calling the
change controls dialog.

3. "Exclude column from file query results" has been added to the
grid edit dialog.

4. The Edit Grid Range dialog has been changed so that the columns in
the results file can be set to be included or excluded.

5. Query > File Query has been changed so that the results file does
not contain columns that have been set to be excluded.

The full version has already been uploaded. If you are a registered
user of the full version please email support@... for the
upgrade details.

The trial version will be available later today.

Steve

I've put npsl1 on Twitter.

http://twitter.com/npsl1

Steve