Multi-dim input in Elman Net

Hi~ I used newelm(simple recurrent network) to train the network. But my input data is a 2-d vector on each time-step , not just a single number. So, how to address this problem ? My code seems wrong:

The input has 3 sequences, and each sequence have 5 time-steps. And again each time-step is a 2-d vector. The output is similar.

 % code
 P = [ [1 1] [2 2] [3 3] [4 4] [5 5];
       [2 2] [3 3] [4 4] [5 5] [6 6];
       [3 3] [4 4] [5 5] [6 6] [7 7]; ]';

 T = [ [2 2] [3 3] [4 4];
       [3 3] [4 4] [5 5];
       [4 4] [5 5] [6 6]; ]';   

 th = [0 10; 0 10; 0 10; 0 10; 0 10;]; 
 net = newelm(th, [5, 3], {'tansig', 'purelin'});

 view(net);

 net.trainParam.epochs = 1000;
 net = init(net);
 net = train(net, P, T);
 y = sim(net, P);

[Greg Heath]

I've never seen NN data like that before. What is it suppose to describe??

[Sin]

sorry, I am a freshman on nntool. The original question is like this:

Suppose we have 3 days` weather data, 5 hours per day. The data structure is (temperature, humidity), so it has 2 dimensions on each time point. Given these data, We want to predict next 3 hours situation.

So, how could I model this problem by Recurrent neural network in Matlab?

Many thanks!

[Greg Heath]

Unclear. I hope you mean the next 5 hours situation

2 inputs

5 measurements per day (e.g., every 4hrs and 48 min) for 3 days

 size(input)  = [ 2 15 ]
 size(target) = [ 1 15 ]

?

[Sin]

this is the data

[Greg Heath]

You will need many days of training data to design a reasonable predictor.

How much data do you have?

[Sin]

5 years data or more

Could you teach me how to coding the input matrix P ? Thanks a lot!

[Meet Shah]

Any answers yet?

I am facing a similar problem