Function change in NARX NN for time series prediction

Hello, I am doing a NARX prediction model by the tool in matlab. I normalized my inputs and targets by standard deviation 1 and mean 0. The values are now both negative and positive( z score). My understanding is I have sigmoid function as default transfer function (Hidden layer) and it’s range is 0 to 1. Now as I have both negetive and positive values I feel I need to change my hidden layer transfer function that can take both negetive values and positive value range (-1,1). I was just wondering wondering how can I change that. The simple script developed by the tool does not contain the option to change The function . Please suggest.

Thanks

Answer 1

[Greg Heath]

>Hello, I am doing a NARX prediction model by the tool in matlab. I normalized my inputs and targets by standard deviation 1 and mean 0. The values are now both negative and positive( z score). My understanding is I have sigmoid function as default transfer function (Hidden layer) and it’s range is 0 to 1.

No. It is tansig = tanh which is bipolar.

The simple scripts in the documentation

 help narx

 and

 doc narx

are worth pursuing before attempting your own problem.

I have posted additional help in both NEWSGROUP and ANSWERS.

 greg narx

 greg narxnet

should suffice.

Hope this helps.

Thank you for formally accepting my answer

Greg