Mahboobeh, I think Maxwell ask about using AdaLiN in MATLAB.

Maxwell, I worked with neuronets in MATLAB about 5 years ago.I use manuals and examples on this extension. Now I don't remember the details, but it seems to me not hard to learn using neuronets by manuals and examples in help resources.

nntool and nftool can be used for this task.

Hello Max;

You can divide the data into a) Training (70%) b) Testing (15%) and c) Validation (15%) using Neural network tool in matlab. So that you can test the trained data using testing runs. 

You can see coding for dividing the data and then you can get the plots of testing.

Sorry members, i meant nonlinear prediction. I have used nntool and nftool to train the network with historical average river discharge. I have developed an ann to predict the following month's discharge. The problem now is how to test the trained network with the sim function using sample data?. Thanks all for the good responses.

In general for prediction problems, You can use cross-validation technique based on a random division of the initial set of data in 3 subsets: for training (weights adjustment), for learning process control (validation) and for evaluation of the quality of approximation (testing). The quality of the approximation can be evaluated by a few measures. the most common is MSE.

Mean Squared Error (MSE) is calculated the difference between to correct outputs and those provided by the network; the approximation is better if MSE is smaller.

you just have to simulate your test data and calculate the MSE or other measures.

Your problem is time series and you can use time delayed networks or NARX in the case you  want to use exogenous inputs in your modelling.

I am searching for the same. Do you find any help?

How to use the trained network for prediction using new input. 

I am working on Feedforward, Elman, and layer recurrent networks respectively.

For instance, in the case of FEEDFORWARD NETWORK,

This code as given in MATLAB worked fine:

[x,t] = simplefit_dataset;

net = feedforwardnet(10);

net = train(net,x,t);

view(net)

y = net(x);

perf = perform(net,y,t)

Thus, the network is trained and ready to be used for prediction. 

My Question:

How can I use the trained network to predict the OUTPUT of say new INPUT (0.0118).

Thank you.

choose build Matlab function

[output] = myNeuralNetworkFunction(input features )

put this line in the output

y1 = a2;

y1 =y1 >.5;

so that the output will be with logical value

from your predefinition if it is

0 --> class1

1 --> class 2

if you are planning to predict your data use:

pnewn=mapminmax('apply', 'pnew', ps)

anewn=(sim(net,pnewn);

anew=mapminmax('reverse', anewn, ts);

note; pnew is new set of data

Training the neural network requires the use of the simulate function (sim).

First, if you have normalized your data before training, you need to do same for the testing data using one of the normalization algorithms like (mapminmax or mapstd etc.)

Second, simulating the network with the new set of data requires you calling the saved network (if you have saved it before). You can do so by using the command "load NetworkName".

Third, to simulate, then use the following syntax,

[output]= sim(net, TestData);

The output can the un-normalized. using the previous un-normalization function coupled with the structure for the output data.

NB: All these are to be done in the MATLAB environment though.

I hope this helps.