How to built time delay network for time series prediction ?

Hi all

I have input p 5x101 cell and target t 1x101 cell. I want to use timedelaynet function to predict this time series. I write this script:

net = timedelaynet(1:8,10);

[Ps,Pi,Ai,Ts] = preparets(net,p,t);

The following error occurred:

Error using preparets (line 161) The number of input signals does not match network's non-feedback inputs.

Error in TimeSeriesPreduction (line 9)

[Ps,Pi,Ai,Ts] = preparets(net,p,t);

[pradeep]

Hi every one,

                 Can any one  please guide me how to divide and prepare input and target cell from the raw data... can u please tell me what is the procedure for time series prediction... starting from beginning...??

please help me out...

Answer 1

[Greg Heath]

The created net does not know it is going to have a 5-dimensional input.

view(net)

Therefore, tell it by using CONFIGURE (help/doc) .

view(net) % Checking

Hope this helps.

Thank you for formally accepting my answer

Greg

[Lelde]

Hello,

I have the same problem. I have 5x1978 input matrix, and 1x1978 output matrix. I tried CONFIGURE, but still receive error

EDU>> net=narxnet; EDU>> [xs,xi,Ai,ts] = preparets(net,x,{},t); Error using preparets (line 161) The number of input signals does not match network's non-feedback inputs.

EDU>> net = configure(net,x,t); Error using network/configure (line 111) The numbers of input signals and networks inputs do not match.

What should I change? Thank you Lelde

[Greg Heath]

%Check size and class (cell) of x and t

 whos x t

Hope this helps.

Greg

[Lelde]

I tried,but data already has the class - cell. What else I can try?

EDU>> whos x t Name Size Bytes Class Attributes

t         1x1978            134504  cell               
x         5x1978            672520  cell

Thank you, Lelde

[Lelde]

I found a solution.

I imported raw data as matrix 5x1978 (x) and 1x1978 (t)

and then used EDU>> x = con2seq(x); EDU>> t = con2seq(t);

to convert data to cells.

Lelde

[Greg Heath]

Good. That was my next suggestion.

Greg

[Mirko Job]

And what about if I have multiple time sequences on which i want to train my network?

[Xunbi Ji]

Maybe this can help https://www.mathworks.com/help/deeplearning/ug/multiple-sequences-with-dynamic-neural-networks.html