How can I get outbounds/uncertainties with Neural Network ?

Hello , I was wondering How can I get the uncertainties bounds from the neural network, I am using simple fitting neural network with (fitnet), I wanted my results to be with sigma and mean rather then just one value. Does anyone know how can I get the results with uncertainties by using Neural Network in Matlab ? I know Baysian Regrassion will proivde that no problem, how to get that sort of reuslts with neural network.

Thanks

Answer 1

[Greg Heath]

Assume that errors are normal with zero mean

 output = target +/- sqrt( mse(target-output))

Hope this helps

Thank you for formally accepting my answer

Greg

[Faisal Khan]

That will give me bounds but those bound will not be true uncertainty. I am looking for bounds which reflect the uncertainties of the data corresponding to the Neural Network results. I Hope you know what I meant.

[Greg Heath]

No. I don't.

Unless you just want to obtain error histograms.

Greg