SwingNN version 2.0b beta 2 is available for testing. The changes are as follows: 1. The problem that resulted in some forecasts going out of range and ...
SwingNN version 2.0b is now available. The changes are as follows: 1. Added check to "Stop when risk is |...|" to Swing Forecast dialog. 2. Added extra...
Ian has been enhancing the network editor so that nodes can be modified and the connections frozen. His additions have shown up a few bugs that could cause...
Hello all, I'd like to EasyNN to recognize and classify objects by a large range of properties. A training row has 200 cols, each an integer value (usually...
... To train a network to identify objects of class A you need some examples of class A and some that are not class A will be needed. It is not possible to...
I imported a dataset with a little over 65000 columns (and only 18 rows) and the import went fine. However, as soon as the data is displayed in the grid (with...
... That's a lot of columns. Learning would be very slow but there is no reason why it should be slow before learning starts. If you send a copy of the data to...
... I don't know how you imported the data as I have spent a few hours trying. The huge number of columns is a major problem. EasyNN-plus is optimised to...
EasyNN-plus 10.0a beta 1 is available for testing. Changes: 1. The toolbars have been changed to Main, Tools, View, Advanced, Reconfigure & Extras. This is to...
EasyNN-plus V10.0a is now available. The changes are as follows: 1. Changed the toolbars to Main, Tools, View, Advanced, Reconfigure & Extras. 2. Added...
Hi, I'm trying to train Easynn with the next simple setup in order to make the network discriminate between the English and German language: two columns with ...
... 80% is higher than I would have expected. I would have thought it is very difficult for a neural network to learn how to differentiate the two languages. ...
EasyNN-plus 10.0b is now available. Changes: 1. Fixed a problem with the temporary version producing inconsistent results. 2. Fixed a problem with multiple...
Hi, I imported the file with the contents below. Importing and making the network went ok, but making a specific Query did not work. If I tried to put 'ad' as...
Hi, If I import a file with a series of training rows and a series of validation rows, the menu option should help to set the imported validation rows to...
Hi, Importing a file with a number of Training rows follwed by a number of Validating rows messes up the data. See the example below. The row with the data: ...
... Hello, Is it producing the "Possible Text Error" dialog? If so, there is a problem in that the error can't be just ignored even though this is not really...
Hi Steve, It is not "producing the "Possible Text Error" dialog" You can use the example-input I gave for reproducing the bug and for debugging. Just replace...
... That's fixed in the next update. ... Most imported files do not have labels that could be reliably used to identify the row type. If you set "Ask while...
... That's what I did and I am getting "Possible Text Error". Which version are you using? Steve -- EasyNN-plus. Build Neural Networks. http://www.easynn.com ...
... I have found the problem. It is a performance change that was introduced ages ago. The check for too close hash values was being skipped while importing....
Hi Steve, I'm using version 10.0a (limited tryout). If after the import and making the network as described in the first mail I do: Query > Query.... Click Add...
Hi, The next combination gives the error 'stl start learning failed'. data in lang2aa.txt with contents word1 word2 word3 language T:0 aa ab ac 0 T:1 ad ae af...
... Otto, I have not been able to reproduce that problem. It does not happen in version 10.0c I have sent you 10.0b and it will be interesting to see if the...
Hi, Steve, The bug is still in 10.0b. You should be able to reproduce it, if you do from sratch the steps as described below, using as test the sample-file of...
... Otto, It is useful to know that the bug is still in 10.0b I know now that I'm not going crazy! It doesn't happen in 10.0c The other bugs you found have...