# how can i provide one input vector to some neurons in the hidden layer and another input vector to some other neurons of the same hidden layer in matlab

1 view (last 30 days)
Preeti Prajapati on 25 May 2014
Answered: Greg Heath on 29 May 2014
I want to create a neural network with 2 input vectors.I am using 2 layer neural network with 10 neurons in the hidden layer. But the problem is i have to provide 1st input vector to 1st 5 neurons of the hidden layer and 2nd input vector to next 5 neurons of the same hidden layer. I searched all the properties of the neural network but none solved my problem. Please help me out with this.

Greg Heath on 29 May 2014
WHOOPS!. At one time when weights were initialized at EXACTLY zero, they would not be updated. Now I don't recall if was MATLAB's code that was changed or it was one of my personal codes.
Regardless, it doesn't work here and I apologize for the bum steer.
What you have to do is to eliminate the connections entirely instead of setting the initial weights to zero. That is explained in the documentation under the name "Custom Networks".

Greg Heath on 26 May 2014
After creating a 10-input net, set the appropriate weights of net.IW to zero; You will have to work out the details but the general idea is
net.IW{1:h1 , i1+1:end} = 0;
net.IW{h1+1:end , 1:i1 } = 0;
Hope this helps.
Thank you for formally accepting my answer
Greg
Preeti Prajapati on 29 May 2014
Thank you for this answer. Sir,I am unable to understand the command: net.IW{1:h1 , i1+1:end} = 0. I tried to assign weights to 0 by using the following code: a=[0.2; 0.3; 0.4; 0.1; 0.2; 0; 0; 0; 0; 0]; b=[0; 0; 0; 0; 0; 0.1; 0.1; 0.3; 0.2; 0.2]; IW = [a b]; net.IW{1,1} =IW; But Sir, this doesn't solve my problem as the weights I set 0 are altered when I train the network. I want to connect 1 input to some neurons of the hidden layer and another input to some other neurons of the same layer.