How can i implement a back-propagation time series prediction using NARX in ANN?

Hi I am beginner in artificial neural network and matlab, I want to predict some step ahead of a coordinate, using previous coordinates in a two-dimensional space. Would you help me?

Answer 1

[Greg Heath]

For an I-H-O (input-hidden-output) net

[ I N ] = size(input)

[ O N ] = size(target)

Thank you for formally accepting my answer

Greg

[roudabe]

sorry would you please it with more details?

Answer 2

[Greg Heath]

1. Read the help/doc timedelaynnet, narnet and narxnet documentation

2. Duplicate and investigate the examples.

3. Decide which one fits your needs.

4. Apply to your data.

5. If you have problems, post with relevant code and error messages.

Hope this helps.

Greg

PS: If you need more example data:

   help nndatasets

Thank you for formally accepting my answer

Greg

[roudabe]

When I want to use narx (by GUI) for this input() and output() matrix the error is :Inputs and targets have different numbers of samples.Please check the pictures I have produced to make it easy for you.

general input is:[x1 y1;x2 y2;...;x3000 y3000] and data matrix is:[x1 y1 x2 y2 x3 y3 x4 y4; x2 y2 x3 y3 x4 y4 x5 y5;x3 y3 x4 y4 x5 y5 x6 y6; ....;x2997 y2997 x2998 y2998 x2999 y2999 x3000 y3000]

and i want to import 6 first column as input and 2 last column as target to train the network

My data generate like this:

   x=linspace(1,20,3000);
  %noise
    sigma=0.05;

%draw a line(path of obstacles) with noise
  y=4*x+1+sigma*randn(size(x));  plot(x,y);
  grid on;

    data=[x' y']; %(3000*2) 
    %%%----> 
    nndata=zeros(2989,12);
    for i=1:2989
        nndata(i,:)=[data(i+0,1) data(i+0,2) data(i+1,1)...
            data(i+1,2) data(i+2,1) data(i+2,2) data(i+3,1) data(i+3,2)];
    end
    inputs=nndata(:,1:6);
    targets=nndata(:,7:8);

Tell me if my way is false to generate data, I want to predict next position of my obstacles which is moving. And I'm not sure about using narxnet or timedelaynet too :( Thanks so much!

[Greg Heath]

See my original answer w.r.t. I/O matrix dimensions.