Click Here for Latest Version and Updates

Frequently Asked Questions

I’m a registered user but I don’t understand the update policy.  Are all updates free?
Minor updates and bug fixes are indicated by the letter after the version number.  They are free to registered users of that version while it is available.  Major upgrades are indicated by a change in the version number.  All major upgrades within three months of the most recent purchase are free.  See upgrades for more details.
How do I change the temporary version to the full version?
Download the full version from the web location that was sent with your registration details when your order was processed.  If you did not get your registration details they were probably intercepted by a badly designed anti spam service or application.  Send a message to upgrade using a valid e-mail address that will allow an answer to get through.
I can't make Query work. What could be wrong?
You are using the Free version.  Querying is available using Q: rows in the Grid.  The more advanced querying facilities are not available in the Free version.
Why does learning in EasyNN-plus use nearly all the CPU?
This is not just while learning, many of the functions in EasyNN-plus use all the available CPU power.  In EasyNN-plus both the application and the threads that need power use whatever is available but they run at the lowest priority.  This allows EasyNN-plus to take full advantage of single and multi processor systems without slowing any other applications.  The operating system and other applications do not slow down.
When I save the Example row names and errors, the values that are saved do not agree with the values that are displayed using the View - Example Errors. Can this discrepancy be fixed?
The errors saved are the differences between the scaled real outputs and the scaled network outputs.  The errors displayed are those that are backpropagated during the learning process.
How do I join the discussion group?
You can join the EasyNN discussion group by clicking here.
How do I get support for EasyNN-plus?
Support for all versions of EasyNN-plus is available by e-mail from support or just click the yellow envelope in the toolbar.
Is an EasyNN-plus guidebook available?
The EasyNN-plus documentation is in the User Manual or this FAQ.  The FAQ is updated with answers to support questions.  The User Manual is updated with every release.
How do I enter my registration number?
When you order EasyNN-plus the keycode and registration instructions will appear immediately on the order screen.  Follow the instructions to enter your name and keycode into EasyNN-plus.  The 30 day trial version will change into the temporary version with no grid size limits.  The location of latest full version with all the extra facilities will be sent by e-mail when your order is processed.
I downloaded the latest version and installed it.  It did not replace the version I already had.  How do I install the latest version?
This is probably because you are reinstalling an old version by running a copy of an old installer program.  You may have a copy of an old installer in your download folder, in your “Temporary Internet Files” folder or in your ISP’s download cache.  Delete all the old copies of the installer and uninstall EasyNN-plus.  Run the installer from the EasyNN web site.  If you still cannot set up the latest version e-mail support for a renamed installer that will bypass any old or cached copies.
I have been running a beta release that has stopped working.  What do I do now?
Follow the instructions that the beta release produced when it expired.  If you cannot remember what the instructions said, load a second instance of EasyNN-plus to recreate the message.
How do I extract the formula that produces the outputs from the inputs?
EasyNN-plus cannot produce a general formula for all networks.  In the full version, from 9.0a onwards, the individual values and formulae used in every node can be exported to a text or CSV file.  See the Excel samples (Add.xls & Test1.xls) for methods of using the formulae. 
Which learning algorithm does EasyNN-plus use?
EasyNN-plus uses backpropagation.  The differences from most other backpropagation based applications are the data structures and the way the data is presented to the learning algorithm.  EasyNN-plus uses double linked lists to store the examples, the nodes and the connections.  The lists can be processed quickly in both directions simultaneously.  The lists can also be extended and contracted dynamically. 
How does EasyNN-plus scale the data?
All inputs and outputs are scaled and offset to the range 0.0 - 1.0  Every column holds separate scaling and offset values so that the original data ranges can be restored.  Some views show the values as 0.0 - 1.0
Do I need to scale the training data?
EasyNN-plus does not require any data to be scaled.
How many training examples and input/output nodes can EasyNN-plus use?
The full version of EasyNN-plus has no example row limits.  Windows 95, 98 & ME can only display about 1700 example rows but EasyNN-plus will still use all the rows.  Windows NT, 2000, XP and Vista can display any number of rows.  Grids and networks with more than 1,000,000 example rows have been created, trained and used successfully.

The number of input/output columns is limited to 10,000 if automatic naming is used.  That is #0 to #9999.  Networks with many inputs and outputs may learn slowly because of the number of connection weights that need to be processed.  

How do I get file queries to work?
'Query from file' is one of the extra facilities available in the full version.
Another of my neural network applications uses different names for the type of examples.  Why is this?
Some applications use the term 'Testing' for 'Validating' while some others use 'Testing' or 'Interrogating' for 'Querying'.  Some others use 'Recall' for 'Querying' and ‘Design’ or ‘Encoding’ for 'Training'.  It can be confusing.
Can the EasyNN-plus engine be used in other applications?
EasyNN-plus does not have an engine.  This is because most of the performance critical code is “inline” to avoid excessive and repetitive function calls.  The weights and biases can be used in other applications to simulate the Querying actions of EasyNN-plus.  See the sample projects that use Excel.  EasyNN-plus can be driven using scripts generated by other applications. 
Which transfer function does EasyNN-plus use?
EasyNN-plus uses the logistic function.  That is 1.0 / (1.0 + e (-net input)). 
How does EasyNN-plus handle missing values?
By default EasyNN-plus substitutes the mid range value.  This can be changed to high, low or average for numeric data.  See Tools > Options > Missing data…
I have a grid with 110 columns and over 3000 rows which needs a few weekly updates.  It takes too long to find the columns and rows so that values can be changed.  The Find command does not work with column and row names.  How do I find the columns that need updating?
Edit > Find works with numbers, words and image names within the Grid.  What you need to use is Edit > Goto.  You can use any text in Goto to search the column or row names.  If you enter numbers then Goto treats them as column or row indexes. Non numeric text or numbers in quotes result in a row or column name search. F4 repeats the Goto.
How can I see how close the estimated outputs are to the real outputs during training?
You need to duplicate the Training examples with Edit > Duplicate Example Rows.  Any types of example rows can be duplicated.  The duplicate row is positioned immediately below the original row and the example type is set to D: for Duplicate.  Duplicates behave as Querying examples.  The estimated outputs are updated by every refresh and can be compared with the real outputs before, during and after training. 
Why do the peaks on the learning graph keep changing?
Up to 5000 graph points are recorded.  This is sufficient for over 200,000,000 learning cycles.  The graph is produced by sampling these points.  The horizontal axis is non linear to allow the whole learning progress to be displayed regardless of the number of cycles executed.  As more cycles are executed the graph is squashed to the left.  This means that the samples of the graph points are not identical every time that the graph is updated.  Some short term peaks may miss the sampling and appear to change.  This sampling is related to the display performance.  It would take a long time to redraw the graph and include every one of the 5000 points. 
How do I change the fragments in the grid?
Fragments never need to be changed because they are regenerated from the images whenever they are needed.  Fragment names in Querying rows can be stepped through in alphabetic order in the Query dialog using the Increase and Decrease buttons.  Changes are written back to the Grid.  The actual name should not be changed by editing and using the Set button.  It will be replaced by unknown ‘?’ in the Grid.
Can an image have more than one name?
Images can have many names.  A warning is produced when a name references an image that already has another name but it can be ignored if needed.  Duplicate images in the same column will have slightly different code combinations. 
If I run a script on my network it changes to a new file.  How do I keep the old file name?
That is how it is supposed to work.  If you open a file within a script it loads a copy of the data into an ‘untitled’ grid as a new file.  That way the original file cannot be corrupted by using a script.
I am producing a script to be used in teaching neural networks.  Can I add some comments to my script?
Any line starting with a non alphabetic character can be used as a comment.
Why do I get the “All learning is stopped” message before my network stops learning?
The learning process is not real time.  The progress message is sent by any learning thread.  The message shown is refreshed by the main thread about once every second when interrupted by a learning thread but will wait longer than a second if needed.  The displays are refreshed from within the thread handling the currently active view once every five seconds.  Threads are not synchronized.  You could get the message five seconds before the view gets a final refresh.  On the other hand, if your mix of networks includes one that takes longer than five seconds to cycle, learning may appear to stop before the message is updated.
Why do I get some hard disc activity while learning is in progress?
The disc activity is not related to the network learning.  EasyNN-plus does not access the disc while learning unless AutoSave or tracing is switched on.  Even then the accesses will be very infrequent.  AutoSave writes a copy of the *.tvq file to disc whenever it is found to be an improvement on the previous copy.  Tracing writes to disc when its buffer is full.  Press F8 to make sure tracing is switched off.  If AutoSave is on you will be asked for the name of the file to use when learning starts.  If you cancel this dialog AutoSave will switch off. 
I am having difficulty navigating around a big grid.  The Home & End keys do not always work the same way.
The Home and End keys move in the same direction as the previous move with the arrow keys.  If you were moving vertically then Home will move to the first row and End will move to the last row.   If you were moving horizontally then Home will move to the first column and End will move to the last column.  You can use Goto to move to any row or column.
I want to create a script that starts learning, waits for 10 seconds and then stops.  No matter what I do it only runs for about a second.  I have tried different waits but it’s always the same.  Why?
The most likely reason is that you are using w(10) command in the script. This means wait for 10 milliseconds not 10 seconds.  See the description of the wait command in the User Manual. The apparent run for “about a second” is because of the time involved stopping the learning thread.  After learning has started a script command of w(0) will result in a few learning cycles because the thread cannot be stopped instantly.  A w(10000) will appear to wait for about 11 seconds. It will wait for 10 seconds and then stop the learning thread which will take about one more second.
I need to create a neural network from my database.  It will need a grid with 17 inputs, 4 outputs and 4,100 examples.  I want it to run on an old 64MB PC.  How much memory will my network need?
The memory requirements cannot be calculated for your network because the number of hidden nodes is not known.  Even when the number of hidden nodes is known it is still difficult to predict how much memory will be needed during learning.  Memory is taken as it is required and released when it is not required.  A slightly bigger network than the one you describe has been monitored.  The initially requirement was just over 10MB.  While learning this grew to nearly 12MB and when learning finished the memory needs dropped back to just over 10MB.  If many other applications are in use at the same time the available memory may start to get low.  It is left up to the operating system to allocate memory.  If memory runs very low while using EasyNN-plus with Windows NT, 2000, XP or Vista you will find that application performance starts to suffer.  With Windows 95, 98 and ME EasyNN-plus could have more serious problems and it may be necessary to close some other applications before restarting learning from the latest network backup copy.
My validating examples are not in the range I set.  How do I find out which outputs are the outliers?
Use Edit > Duplicate Example Rows… on the validating rows.  You can then compare the desired values with the calculated outputs before, during and after learning.  The outliers will usually be obvious.  Alternatively Action > Check Grid will find any validation rows outside the training range. Edit > Exclude Outside Limits and Edit > Delete Excluded Rows can also be used to remove any type of out of range rows.
When I import a file, I get an error saying, “The line read operation has been terminated after 10000 characters without a line end".  How can I import the file?
With no line feed characters, file reading stops after 10000 characters.  If the file is really text then insert line feeds at the points were the lines end.  The line feeds are used to delimit the example rows.  One of the first things that the import does is count the lines so that it knows how many example rows to create in the grid.  The limit for text files is raised to 1000000 from version 5 onwards.  Binary files have no limit.
How does EasyNN-plus know when to add hidden nodes while it is learning?
The average error is saved every second.  The last saved error is compared with the current error over a few cycles so that the error trend can be established.  If the trend shows an increasing error, a node is added.
My file takes over a minute to import but only produces 304 rows. Why does it take so long and what happens to the other rows?
There are many checks on the file before the import starts.  Some of these checks are very time consuming.  You then select the line on which to base the number of columns or column headings.  All the other imported lines must then fit into the grid.  The dialogs created during the import explain what is happening but you may have taken the option to ignore the rejected lines.  The last dialog is a summary of what has happened during the import.
What are Clusters for?
Clusters show related columns in data where no obvious inputs and outputs exist.  For example:  Census data will have no inputs or outputs but by finding Clusters some relationships will be found.  The columns in such Clusters could then be set to build networks. 
Is there a way I can use information from 'Node and Connection Details' or from the output of special files to be able to refit the same model without having to start from scratch, i.e. waiting for the model to be rebuilt from the first cycle to the desired number of cycles.
The network is not rebuilt.  The files are created by a snapshot of the current state.  If learning is finished the snapshot will not change.  The details are a snapshot on exit.  Special files can be snapshot at any time.  Using a script the network can be snapshot while learning.
When I start my network learning I get a backup file written.  I don’t want a backup every time so I checked the “do not ask for AutoSave settings when restarting learning”.  It still produces a backup.  How do I stop it now that I don’t get asked?
It is only the automatic “ask” that is inhibited.  Options > AutoSave will still create the dialog so you can switch backup off and on as required.
I have a neural network that is used to predict wave motion.  Measurements are added every day.  It all works very well when the new examples are imported but if a script is used to automate the import I get a failure.  What could be wrong?
The script is not setting the row types.  With an import the row types are requested during the initial dialogs.  When a script is used the row type must be set using a rt command.  Alternatively record a macro to use for the daily import.
How do I convert a column with two words like “high” and “low” to a bool?  Switching from text to bool changes both words to true but the text “low” needs to be false.
No direct conversion from text to bool is provided but it can be achieved.  First switch to integer.  This will give the encoded text values.  Replace the value that needs to be set to true with 1 and the other value with 0.  Switch to bool.
I get an investment file from a person using EasyNN.  EasyNN-plus reads it but also converts it.  How do we exchange files?
EasyNN-plus has many file components that are not used by EasyNN.  It adds the framework of these extra components as it converts the file.  If you need to exchange files between the two applications you must answer no when EasyNN-plus prompts to convert the file.  EasyNN-plus can use EasyNN files without conversion but EasyNN cannot read EasyNN-plus files if they have been converted.

Your associate should upgrade to EasyNN-plus if the extra facilities are needed.  See upgrade for details.

Why is the message I sent to support being ignored?
Support messages are never ignored but may not have been seen.  The usual cause is that the message is in HTML and so it has been filtered out as SPAM.  Support messages must be in plain text. 
I upgraded from EasyNN to EasyNN-plus and wanted to see if an old network produced the same results in both versions.  When I try to load EasyNN it crashes but EasyNN-plus works OK.
Once EasyNN-plus has been run it sets up its screen and file status in the registry.  This is not compatible with EasyNN.  If you are confident with regedit then you can make EasyNN work again temporarily by deleting the registry entry:- HKEY_CURRENT_USER\Software\EasyNN\EasyNN\EasyNN-Summary
The 30 day trial period will end just before I finish my project.  Can I have a few more days?
The day count can be changed to extend the trial period by five days but the other restrictions and limits cannot be removed because the enabling code is not included in the trial version.  Send an e-mail to support for instructions.
Do I have to pay VAT when I order EasyNN-plus?
If you order from a country in the European Union you will be charged VAT by the order processing company.  Orders from other countries are free of this tax.

With multiple outputs how does validating combine the outputs?

If Controls are used all outputs must satisfy the conditions of the validating test.  If Rules are used then each output can have different validating tests.  Controls and Rules can be combined to produce very flexible validating tests.

My grid has thousands of fragments. Why does it use less memory when it is reloaded?

The fragments are held in slave locations until they are no longer needed.  When images are saved, copied or moved the slaves are checked for images or fragments that are not in use.  Any unused fragments are cleared.  The image slaves can be checked for free slots using Tidy Text and Images.  Text slaves are checked at the same time.

Why is the target error in the tvq file much higher than the target that has been set in the controls?
The tvq file mapping is controlled by the file status.  This can be seen with tracing switched on.  The original tvq file had 10 free locations for extra variables as development proceeded.  These locations were used up a long time ago.  The file format now used can be extended as needed and does not require any free locations.  A variable is needed to differentiate between the two file formats.  Targets above 1000 indicate the new format.  They are adjusted to the actual target when the file is loaded.  Old format files are automatically converted to the new format.
How do I get new rows to use the missing data option?
New rows do use the missing data option but the values are set to ‘?’ because they are still unknown.   If you want to create the new row with the actual missing data values use Insert.  Rows can be inserted anywhere in the grid including the start and end.
Why does tracing try to start when I open a file?
You are using a beta version.  Get a copy of the latest release version.
My neural network runs continually with AutoSave set to save the best validating results.  It also adds hidden nodes while training is progressing.  After running for a few hours I checked the best file saved.  It still had the original number of hidden nodes. Why were the other networks not saved?
The original network was producing the best results.  No amount of training or adding extra nodes produced better results and so the original network was never overwritten.  It is quite common to find that small networks produce better results than large ones.
Is there an easy way to change columns when looking at the column values graph?  It is not easy to keep switching back to the grid.
The Grid view navigating keys and the Goto dialog will still work while using the column values graph view. 
What is 300000000549775580000000000000000000000.0000?
3e38 is the number used to indicate an unknown value in a Grid cell.  It should only be seen in the trace file and Macro files.
Where can I find a crack for the full version of EasyNN-plus?
Is it really worth the risk?  Downloading from "cracks" and "warez" sites is not "safe computing".  Some crack sites offer a "registered" version of the software that is actually a "trojan horse".  Such programs are not always detected by anti-virus programs since, after all, they are not viruses.  Yet they can let somebody else take over your computer, transmit or corrupt your files, or even erase your entire hard disk.
What is “eos” for?
It is a script command that marks the end of the script.  The scripts will stop when the end of the file is reached but eos closes the script tidily.
Why do the predictions jump up and down while the network is learning?
The vertical position of the dots is determined by the output of the network for each example row.  If the learning rate or momentum is high the predicted outputs will oscillate.  While learning these oscillating outputs will be clearly visible on querying or duplicate rows.  As learning progresses or if decay is checked the oscillation should be brought under control.  If the oscillation continues the learning rate needs to be reduced.
I know about T, V & Q but what are the other prefix letters for?
Prefixes are used to identify the types of rows and columns.   The ?: button switches prefixes off and on in the grid.  T: is a training row.  V: is a validating row.  Q: is a querying row.  D: is a duplicated row.  F: is a forecast row.  E: is an engaged row.  P: is a prediction row.  I: is an input column.  O: is an output column.  S: is a serial column.   X: is an excluded row or column. 
Why does Zoom Wider not work?
The Wider and Narrower zoom controls are enabled when the Network view is selected.
How do the conditional functions work?
A sample function is described in the right click Set Function dialog.  The sample ifz +1 x5 function gets the value from the grid, if the value is zero the function adds 1 and then multiplies the result by 5.  Execution of the whole function is from left to right.  If the condition is true execution continues.
Why does my network file take such a long time to open?
To save time later nodes are initialized while the file is opening.  Huge networks take a long time to initialize if your network has thousands of nodes.  In version 6.0a onwards all networks only take a few seconds to open and initialize.
When I import a file the last line is ignored.  How do I include the last line?
In EasyNN-plus versions before 6.0a the last line of the file can be missed if it does not end with a line feed.
No matter how long I leave my network learning, the average error stays around 0.5 and never gets much lower. What could be wrong?
It is probably because you have some conflicting examples.  Use Check Grid to find and remove the conflicts.
How do I use the query arrow buttons when on the network view?
The query toolbar arrows are enabled when a querying row is selected.  Select a querying row in the grid before switching to the network view.
Remove Duplicated Examples Rows completely misses identical rows.  Is this a bug or am I using it incorrectly?
You should be using Action > Check Example Conflicts.  Duplicated examples are those produced for testing purposes by Edit > Duplicate Example Rows…
How do I make file queries work with serial mode columns?
This problem is caused by serial mode columns being identified incorrectly when they are counted for the file query.  The bug is fixed in 6.0c
How do I get Find to scan along the row?  It only seems to work in columns.
Edit > Find does not work when it is set to search rows.  The bug is fixed in 6.0d
What are relative rows and columns used for in Set Function?
The cell in the relative position is used for the operand in the function.  For example +g4 will add the contents of the current cell to the contents of the cell 4 rows below it in the grid.  The result is returned to the current cell.  In the above function Execute on Input/Output Column will add the contents of the cell at the selected column, row 0 to that in row 4, then 1 to 5, 2 to 6 and so on.
Why does forecasting loop when than more than one column is set to serial?
The process that detects the maximum cycle period takes a long time and could appear to be a hang up.  The process is much faster from 6.0e onwards.
When I play my macro the last column is not set to output.  What could be wrong with the macro?
This is a deficiency in the column indexing when the macro records a text file import with the first word(s) of the line being used for row names.  The problem fixed in 6.0f
How do I use the correct date format?
The date and time are represented as numbers in columns set to modes integer, real or text.  In binary or image mode columns the date and time are not inserted.  In 6.0f the row and column names is formatted using the system default locale settings for the date and time.
How do I import a file created with Excel?
Save the Excel file as a comma delimited type.  This will save the file with a CSV extension.  EasyNN-plus will import the CSV file.  From 6.0f onwards it is possible to import Excel XLS files directly.
My script ends with showwindow(5) so it should maximize. Why does it stay minimized when it is run from the command line?
The window probably needs to be restored.  If you add a showwindow(10) at the end of the script it will restore.  From 6.0f onwards the window does require to be restored.
When I use Replace to fix a spelling mistake in one word in many different rows the results are not what I expect.  How should it work?
Replacing part of a text string in whole rows or columns is not reliable.  The problem is fixed in 6.0f
Why does my trace file only contain the Seek activity?
The trace file contains the latest 1000 trace entries.  Seek functions produce a thousand trace points very quickly.  Only the last few entries in the trace file are needed for diagnostic purposes.  Tracing can be switched off most of the time.
When a file is opened, why are the validating results different to the ones on the graph?
One refresh cycle has to run to recreate the graph and other views.  This can result in a slight change to the validating results.
I already have a grid.  How do I add more rows by importing files from Excel?
The files with the additional rows need to be in CSV format.  Excel can save files in CSV format.   In this format the low Excel row and column limits are avoided.
How do I run my script from the command line with EasyNN-plus hidden?
At the start of the script add a showwindow(0) or sw(0) command.  The script will then run with EasyNN-plus hidden.  It will remain hidden until another showwindow is executed with a parameter greater than zero.  Scripts and macros that do not make EasyNN-plus visible may leave the process hidden and running.  If the script or macro ends or fails while hidden an attempt is made to become visible.  Do not assume that this will always be successful.  Some script failures cannot be detected by a hidden process.  Windows Task Manager can be used to kill hidden processes.
What is the long string of digits starting with Min in Edit grid?
The digits show that the Min and Max values are not known because nothing has been entered into the column.  They should be shown as an unknown value ‘?’ indicator.  The string is not used and can be ignored.  The problem is fixed in the 6.0i
Why does the Grid appear a few minutes after I have selected the errors view?
If the network is learning the view will switch back to the Grid after 100 refreshes.  This is to improve performance. Learning is suspended while recalculating the graph or bar views but not the grid view.  Clicking on the error will go to the example row in the Grid.
How do I remove query rows when I am using the query dialog?
Query rows can be deleted or excluded from the grid but the query dialog needs to closed first.  From 6.0j the query dialog will have an exclude query button.
How do I avoid the crash when I am trying to select a whole row?
This is a bug in Edit > Select > Row when there is no row to select or when the cursor is positioned in the row label area.  Position the cursor on the row to be selected and then use the right click menu or double click the row label.  The bug is fixed in 6.0j
Why does the day count jump by ten days?
This is a bug in the trial version of 6.0k that can cause the day count to increment by 10 or more days.  Send an email to support for a method to reset the count back to 1 day.  The bug is fixed in 6.0m
What happened to version 6.0l?
Version 6.0l was ready for test but then the problem with the trial version of 6.0k was reported.  The problem was still in 6.0l so another build was needed.  The bug was fixed, version 6.0m was built, tested and released.  Version 6.0l was never released.
Why does replacing column values in my grid take so long?
It is a problem with the very long text strings in the first column.  The editor is treating the whole grid as text.  The problem is fixed in version 7.0a along with a huge improvement in editor performance.
Why is my web site not listed with the other clients?
Personal web sites are not listed.  The list of companies and educational establishments is slowly being developed as sites are found and verified.
Can I help test future developments?
Beta releases of EasyNN-plus are announced on the discussion group.  Join the group by clicking here.
When I save the predictions with the grid they are clearly wrong.  Is this a bug?
Tools > Special Files > Grid... works if predictions are not included.  The problem is a bug in duplicating rows which is fixed in version 7.0c
What happens if I put a script in a macro, then the macro file back in the script and then run the script?
The script that is run will be a series of commands as in a normal script.  The result will depend on the commands.  Scripts can be edited to cause a crash but most will run safely.  Endless loops should not be created and only one loop can be produced with the backto or bt script command.
Why can't I set the open defaults to remember where the cursor was when the file was closed?
The facility will be available in version 7.0h
I don't want edit defaults to move the cursor unless I have changed the value.  How do I set that up?
That is the way the default works in 7.0g.  It is optional in 7.0h 
When running hidden from a macro it won't close, why not?
If EasyNN-plus will not close, some threads have not finished.  The thread warning will be missed if running hidden.  Find the problem before running your macro hidden.  It is usually quite easy to find macro problems but you may need to run the macro as a script after adding a single step command.
Why is Action > Forecast disabled?  My network trains OK.
Forecasting needs one or more columns set to Serial.  Nothing else is needed.  Networks do not need to be trained first.  Forecasting will create and train networks as it needs them.
Why do I get a grid size limit when running the Races part of Getting Started?
An early build of version 7.0 was released with a size restriction.  The restriction has been removed in the current release of version 7.0
My installation keep changing back to the trial and so  have to keep entering my key code.  How do I make the keycode work properly?
You have installed a "cracked" copy.  Cracks will not run for long without problems.  Download the full version from the web location that was sent to you when you bought EasyNN-plus.
What is the vertical dotted line on the column graph?
It shows the position of the selected example row.  If you left click on the line the view will change to the Grid view and show the example row details.  If you click anywhere on the Grid the line will move on the Column view.
How do I detect outliers?
You can use View > Example Errors to find the rows that produce large errors while training.  These rows usually have outlying values.  Edit > Exclude Outside Limits is useful for removing outliers from the training set.
Why is the sensitivity the same on all outputs?
The refresh button needs to be clicked whenever the output is changed.  This is automatic in version 7.0j
I can't make sense of the column graph even if I zoom in on the part I want to see.  How am I supposed to understand the trend with all those noisy lines in the background?
Right click on the graph to get a selection menu.  Any of the graph lines can be disabled and enabled.  Left click to go to the equivalent row position on the Grid.
How do I set the defaults so that I can type numbers directly without having to open the dialog and move to the next cell?
Set Defaults > Edit... to 'Right', 'Only...' and not 'Select...'.  That will allow anything to be entered directly in the grid Edit Cell dialog and then move to the next cell on the right.
Why doesn't the grid scroll when I press the horizontal scroll bar?  Right scroll moves one cell to the right but left scroll doesn't move at all.
This is a bug in the mouse code that must have been introduced with an upgrade.  It's just the mouse left click that doesn't work as it should do. It is fixed in 8.0e
If I use "Stop on score..." it crashes on validate.
This is a problem generating the Graph > Learning Progress view.  It is fixed in the 8.0g update.
Why do the grid rows break up when I use the mouse wheel?
The mouse wheel was disabled in earlier versions but it has become enabled in the current version.  The problem is fixed in the 8.0g update.
Why does the status bar only say 9999 rows when many more rows have been imported?
This is a bug in import for *.txt and *.csv files.  The actual number of rows imported is correct but the report stops counting at 9999.  The report can be safely ignored because the import will have worked.  The bug is fixed in the 8.0h update.
If I use stop on cycle 1000 the progress graph shows 990 when learning stops.
This is only a display problem.  The 1000 learning cycles will have been completed.  The problem is resolved in the 8.0j update.
Why does the free version keep asking me for my name and number?
A recent minor update was uploaded with the wrong controls defined.  The trial and free versions prompted for details as if they were the full version.  The uploaded files were replace as soon as the problem was noticed.  If you are still having trouble download a new copy of the version you want to use.
Why did my trial copy run out on day 2?
This usually happens when a network file is used that has a day count older than the EasyNN-plus application day count.  You may have used an old file that had been created by someone else.  The day count in the network file had never been initialized and so it was set to some huge number.
What are the four range buttons for in the properties?
The buttons were intended for a change in the range extension facilities.  This was withdrawn but the buttons were not deleted.  They are removed in the 8.0r
Why do the selections in the query dialog take so long to work?
Queries can be anywhere in the grid so every row must be checked.  If there are thousands of rows this can take a few seconds or even minutes in versions before 8.0r  The search takes just a fraction of a second in the 8.0r
Where is NNDemo.ppt?
The file was not included in the 8.0q installer build.  It has been added back in the 8.0r update. 
If I run a script which fails it won't let me fix it with the editor.
That is because some scripts can fail in a way that leaves the script file open and connected to the network file.  The problem is fixed in 9.0a
When I add a new bool name a message says that I can't use an existing letter. I delete the bool name that uses the letter but I still get a message about an existing letter.
It is due to a failure to delete the leading character of the name when it has been duplicated. The problem is fixed in 9.0a
My network auto-saves when the validating results improve. The saved file slowly grows. Is this some sort of leak?
There is no leak. The small increase in file size is normal for grids that have validating examples. Every time validating runs the validating errors are saved so that they can be used for drawing the progress graph. The saved file will not grow very much and it will stop growing when the validating errors stop changing.
If I add some new names for my classification output they are deleted when I reload the file. How can I make sure they are saved?
It is a bug in opening files that have new names for boolean values. The old names overwrite the new names. The bug is fixed in 9.0a
What happens if I run a script with pause and restart using another script?
That all depends on what else the scripts do. Switching scripts can be useful. The 'pause' saves the script line count. The restart continues from the next line. If the next line is a RunScriptFile then the second script could run without manually switching to it.
How do I know when a script has been paused?
The pause command has an undocumented parameter.  If you use pause(1) in the script it will ask to continue.  The user documentation is corrected in 9.0a
Why are my notes being deleted when I add a column?
It is because the notes need to be relocated when the cells with notes are shifted left or right.  Inserting, adding, pasting or removing a column to the left of the note will cause a shift.  The notes are not being deleted but they are losing their column reference number.  The problem is fixed in 9.0a
Why does my file opening time vary from two or three seconds to over 10 seconds?
Since upgrading to version 9.0a you will find that the file opening times vary.  This because the results of the row error calculations are held in lists and saved with the file.  The row errors are available as soon as the file is opened so that the error displays do not need to be recalculated.  If you do not need the error displays you can clear the lists with Tools > Tidy Lists.
Why do my text entries in Excel fail to import?
This is a problem with coding the text in *.xls files.  It also slows down the import.  Both these problems are fixed in version 9.0b
When training my network a row limit message appears on the status bar before the limit is reached. Training works OK but the message is wrong. How do I stop it?
It is a spurious status message that only appears in the trial and free versions. It does not cause any problems. The message is removed in 9.0c
I left a big NN learning at night. It was expected to run for hours but when I checked in the morning learning had stopped at 04:13 and the last backup file was 04:03. What could have stopped learning?
If set in AutoSave, backup will run every ten minutes while the network is learning. If anything goes wrong with the process, learning and backup will be inhibited. The backup is written to a work file first, which is then checked and then renamed. The last successful backup file is not overwritten. In this case the 04:13 backup to the work file failed leaving the 04:03 file.
Why does importing a file fail with an access violation?
It seems some editors use \n (newline) for line feeds rather than \r\n (return newline). EasyNN-plus tries the fix the format by replacing \n with \r\n but it doesn't always have write access.  The problem is fixed in 9.0c
Spreadsheets with only one column fail to import.
This is a bug that is fixed in the next update. For now just add an extra dummy column to the xls file. The dummy column can be deleted from EasyNN-plus grid after the import.
I get an "illegal operation" message when I create an unconnected network.
This bug is fixed in version 9.0e
If I delete a node with the network editor the background color changes and then crashes.
This is a bug in the node isolation code that causes problems as the rest of the layer tries to reconnect.  The bug is fixed and the network editor is improved in version 9.0f
Why does my neural network always take different times to learn?
Initialization varies because the weights and node biases are set with random values and so learning is starting with different weights and biases.  This can make a difference to the learning time.
Why are some of the text inputs incorrect on queries?
This is a bug in the recovery of short text strings.  The bug is fixed in version 10.0c
I would like to drop files onto EasyNN-plus. How do I make it work?
'Accept drop' is enabled in version 10.0d tvq files.

If I use a query file the result file always has question marks in the bool mode columns.

This is caused by using non boolean values in boolean columns. Version 10.0g will accept 0 for false and non 0 for true.


Copyright © 2002 - 2008 Neural Planner Software
Last Updated: 30th September 2008