Here is a feedforward, multiconnectivity, Sigmoid transfer function,
backpropagation, multi layer perceptron Artificial Neural Network source code in VBA;
that I needed to make for my studies. It will solve the XOR nonlinear
separation problem and graph out the learning path.
This is a very clean implementation of a neural network in VBA. Actually one of the cleanest I've ever seen. One question though, and I apologize in advance if this is a stupid question, but where are the learning and momentum terms located?
ReplyDeleteHello!
ReplyDeleteI looked through the file today looking for the terms and found an important bug - while calculating other sigmas, they were not accumulated - now ok.
The learning term is the variable "nyy". The momentum term has not yet been implemented.
Looking forward to hearing from you,
Jaan
Jaan,
ReplyDeleteI've created a fairly interesting tool based on your ann code. I would like to have you take a look at it and see what you think as well as if there are any improvements that could potentially be made to it. Is there an email that I can send the workbook to? Please let me know.
Hi! My email address is "jaanbi at staff.ttu.ee".
ReplyDeleteDear Jaan
ReplyDeleteThanks for your knowledge sharing. I try to solve my problem with your code,it's f(x) = sin (x1) + sin (x2)+sin (x3) +sin (x4) with 4-10-1 network.
I put this change to your provided file ant I also change Input and Output range according to my problem. I produce 10 series Input and Output data to train network. I read code line by line and I understand it so-so. I'd like to put a stop criteria according to calculated error ( e=z(1-z)(y-z) ) if it lower than epsilon. I didn't Understand this calculated data show me what also. Please help me.
MyEmail: ansar dot rezaei at gmail
Thank you so much
Ansar