to prove the robustness of neural network model what is the best model which can compare to it ? (especially in order to forecast)

to prove the robustness of neural network model what is the best model which can compare to it ?

can you propose a model?

THANKS

Answer 1

[Greg Heath]

No.

I measure robustness by adding increasing levels of noise to the input.

Hope this helps.

Greg

[coqui]

can you explain more please.

I have five input series to predict only one series(predicted price).

For robustness checking, what I can do?

[Greg Heath]

Which timeseries function are you using? Timedelay, or Narx?

Either way, add noise at a fixed SNR to the input and plot output error vs SNR.

 [I N ] = size(input0)
 var0   = mean(var(input0'))
 input = input0 + sqrt(var0/SNR)*randn(I,N);

Hope this helps.

Thank you for formally accepting my answer

Greg