How to constrain Neural Network output weights to fixed values

Hello everyone,

i have the necessity to train a NN to predict a certain output y(t) given a dataset [x(t) w(t)]. where x and w are signals in time [x(1) x(2)...w(1) w(2)....]

I need the shape of the network to be:

y(k)=F(x(k))+G(x(k))*w(k)

Meaning that i give x and w as input to my network and the training procedure has to learn something like the equation before. F and G can be nonlinear but the relation between G and w has to be G(x)*w and this last product has to sum with F.

By playing with layers of NN i defined this structure in the image attached. My idea is to "fix" weights on the output layer to be 1 and w(k).

How can i do this in matlab? Is it possible to do this?

Thanks in advance

Best

Gianni

Answer 1

[Srivardhan Gadila]

Refer to Weight and Bias Values, Input Weights, Layer Weights & net.layerWeights{i,j}.learn.

You can access the layer weights as follows:

net.LW{i,j}

You can set any values to the above weights and set the net.layerWeights{i,j}.learn to 0 so that the weights won't be altered during the training & adaption. In this case setting a specific weight for a connection is not possible since the property net.layerWeights{i,j}.learn is defined for the entire connections between layers i and j.

net.layerWeights{i,j}.learn = 0

net.LW{i,j} = ones(size(net.LW{i,j})) % any weights of size(net.LW{i,j})

If your network architecture is defined and trained already:

Then you can set weight of a connection between nodes k & l of layers i & j as follows:

net.LW{i,j}(k,l) = 1

and then use the network.

The above things can be done to Input wieghts too.

[Giovanni Domenico Di Girolamo]

Thanks for the answer. Very helpful.