Re-implementing a Neural Network from matlabs Toolbox as an mfile

Hi,

I have a problem where I have trained a 10 neuron single layer network with three inputs and two outputs and gotten good results but need to re implement the evaluation function of the network as a standalone file. That is, I want to take the information stored in the neural network object, weights etc., and hardcode it in to a function.

The naive way (schematicaly described)

outputsFromHidden = activationFunctionHidden(inputWeights*inputs + inputBias) networkOutput = activationFunctionOutput(hiddenWeights*outputsFromHidden + outputBias)

does not work, and I guess that there are some input rescaling, and possibly other things going on in the evaluation function of the NN-object in matlab.

Does anyone know the exact computational flow of the evaluation function (i.e output = net(input) ) and where in the NN object one finds the necessary values?

Best Regards, Robert

Answer 1

[Greg Heath]

Search ANSWERS using

tsettings

Hope this helps.

Thank you for formally accepting my answer

Greg