Author 
Topic: Weight & Bias 

h_amooie 
Posted: 08Nov11 23:12 



Hello all; I have trained a network with a data set by nntool in MATLAB. everything is OK and the result of simmulation for predicting a value beyond the data set is correct and satisfactory. BUT: For some reasons I need to extract the weights and bias to complete my job,..As I try to test the mathmatical combination manually with the weights and bias matrices gotten in nntool I am unable to reach to the correct result. Can anyone tell me the reason????
Tnx 


pavl 
Posted: 09Nov11 03:00 



Hi,
I have a similar question.
I trained a neural network as follows:
networkParameters.epochs=1000; networkParameters.goal=0.00001; networkParameters.max_fail=10; networkParameters.trainRatio=1; networkParameters.valRatio=0; networkParameters.testRatio=0; net=newff(trainingSetX,trainingSetY,hiddenNeronsNo); [net]=train(net,trainingSetX,trainingSetY);
Then I took the weights and biases:
b = net.b IW = net.IW{1,1} LW = net.LW{2,1}
So I have (for the 3 hidden neurons): Three Layer Weights, Three Input Weights 3 biases for hidden layer and a general bias
I want to get input for a given output. y = sim (net, x) (x  a scalar)
But without the use of Matlab. 

