FeedForwardNet's Algorithms

I have a trained feedforward network with 2 hidden layers, and now I want to implement that network on my webpage on which users can input some value and get corresponding answers calculate by the neural network. To do that, I am thinking of a javascript. And I wrote a code in matlab to test the algorithms.
The problem is that the answer calculated directly by the network is totally different from the answer by my code. Would someone tell me where did I go wrong? Or is there an easier way to do that on my webpage?
load(['../result/' savename], 'nnet', 'ntr');
iw = nnet.iw{1,1};
lw1 = nnet.lw{2,1};
lw2 = nnet.lw{3,2};
b1 = nnet.b{1};
b2 = nnet.b{2};
b3 = nnet.b{3};
ymin = nnet.inputs{1,1}.processedRange(:,1);
ymax = nnet.inputs{1,1}.processedRange(:,2);
xmin = nnet.inputs{1,1}.range(:,1);
xmax = nnet.inputs{1,1}.range(:,2);
inp = [1;2;3;4];
inp = (ymax-ymin).*(inp-xmin)./(xmax-xmin) + ymin;
a = tansig(dotprod(iw,inp)+b1);
b = tansig(dotprod(lw1,a)+b2);
c = purelin(dotprod(lw2,b)+b3);
ymin = nnet.outputs{3}.processedRange(:,1);
ymax = nnet.outputs{3}.processedRange(:,2);
xmin = nnet.outputs{3}.range(:,1);
xmax = nnet.outputs{3}.range(:,2);
c = (xmax-xmin).*(c-ymin)./(ymax-ymin) + xmin;
c1 = nnet(inp);

Answers (0)

Categories

Find more on Deep Learning Toolbox in Help Center and File Exchange

Asked:

on 10 Jul 2011

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!