I could ramble on for a while about neural nets and how they work, but I find that his site is best at explaining what is going on.
http://clemens.bytehammer.com/papers/BackProp/index.html


>Hi Grady. Yes I understand this and I did download and try the app before posting. My questions remain - what sort of inputs, what does the program attempt to do with the inputs, what are the outputs, how does it get there from the inputs. iow, you need to explain more about the mechanics that the program is doing with the inputs to generate its outputs.
>
>
>>From a user aspect, it's quite simple.
>>There are three forms. The first allows you to get a table name. The second allow you to select the field to use and the third generate the code for the neural network.
>>
>>There is a working version available for download on the UT and codeplex, but because I can't hold a candle to the talent in the UT, I have placed it on codeplex for improvement and perhaps redesign. The app is free for any VFP developer to use at will. Docs are included in the zip.
>>
>>>Grady
>>>
>>>I think it might be helpful (at least to me) if you explain a little bit about the process involved; what is the objective, how does the program attempt to get to the objective, what the program inputs will be, what the outputs will be. How the inputs are converted to putputs in your example program. But in more detail than the textual description of the crime example on the codeplex.
>>>
>>>>I believe I have managed to get a project started on the codeplex site.
>>>>
>>>>Anyone who has coded a neural network's weights and codes and hasn't gone insane from it is welcome to help out with the project. It is titled, "A Neural Network Code Generator for Visual FoxPro"
>>>>
>>>>I'll warn you now, that my ability to code is . . . well . . . sloppy is a good word to describe it. It is a relatively small app with three forms, but it's output can be huge if you require a lot of nodes for you neural network.
>>>>
>>>>The final product is free to all VFP users. Here's the URL http://www.codeplex.com/nnp
>>>>
>>>>Thanks folks.
>>>>Keep having fun.