Primary key & Rule Extraction
Message #343 -- Wednesday, November 06, 2002 -- 09:51
The rules are embedded into the network weights. I know of no way to extract rules from a neural net. If you want to see the rules, try VisiRex instead of NeuNet Pro.
A primary key field is a database column that contains a unqiue ID for every row. It is usually just a simple row count. In Microsoft Access, you set this by clicking the yellow key icon. If you use NeuNet Pro "import" for text data, a primary key will be automatically created for you.
Rule extraction
Message #342 -- Wednesday, November 06, 2002 -- 08:18
Related Messages:
J said:
Is there any way to extract rules out of a trained neural net using NeuNet pro?
(No Subject)
Message #341 -- Wednesday, November 06, 2002 -- 06:31
Related Messages:
HI WANT TO KNOW THE MEANING OF PRIMERY KEY IN INFORMATION TECHNOLOGY.THANK YOU
about other data
Message #340 -- Friday, November 01, 2002 -- 01:28
Related Messages:
I can't run ir with own data.Only can do this program with the example data.But you said level one is free.How can I do?
Number of samples for modeling
Message #338 -- Wednesday, October 16, 2002 -- 03:30
Related Messages:
Hi,Anybody knows how many samples needed for modeling. T
Back Prop
Message #337 -- Friday, October 11, 2002 -- 09:29
Hi Tim. It uses the classic backprop algorithm with sigmoic activation. It is a feed forwards/backwards propogation net.
Back Prop
Message #336 -- Thursday, October 10, 2002 -- 18:40
Related Messages:
Does the Back Prop neuro net use feed forward? In other words, is it a feed forward/backwards propogation net? Thanks!
installation
Message #335 -- Friday, September 27, 2002 -- 09:30
Our programs will install onto any 32 bit version of Windows (i.e. higher than version 3.1)
installation
Message #334 -- Friday, September 27, 2002 -- 07:28
Related Messages:
hi,
would like to know if the neuNet software is valid for windows Me
thanks
When to Stop Training
Message #333 -- Friday, September 06, 2002 -- 10:16
This is a hit or miss process. Normally you want to train until the training error will not improve any more. If fact this may be TOO MUCH training and the performance on the unseen test set will be worse. It depends how clean your data is and how well shuffled between train and test sets.
If you are doing anomaly detection, you want to leave some error. The amount of error you leave defines the threshold of your anomaly.
NeuNet Home