sequence to sequence, working in batches for LSTM and/or NLARX

Per batch, I have two input sequences and one output sequence. All three have the same (limited) length for training (in this example 30). Every individual batch has the same length (in this case 30)

So each batch looks like (in case of length of 30):

Input 1: X1-1, X1-2, X1-3.......X1-30

Input 2: X2-1, X2-2, X2-3.......X2-30

Output 3: Y1-1, Y1-2, Y1-3.......Y1-30

There will be a x amount of batches with the same builtup. Each batch need to train the LSTM and NLARX system/network until the network is trained for all batches. So I assume the batch size will be one.

After the training of the network the goal is to forecast for one batch the Output Y given:

Input 1 (1:30), Input 2(1:30, and part of the Output. For example Y(1:15) and then predict Y(16:30) or a variant of the division (1:20 and 21:30).

As I am not proficient enough with deeplearning, can someone help with the requirded code for using LSTM and for using NLARX, for training and forecasting per batch as described above, that would be of great help?

Many thanks in advance.

Best wishes

Leon

Answer 1

[Venu]

Hi @Leon

You might find these documentations helpful

https://www.mathworks.com/help/deeplearning/ug/sequence-to-sequence-regression-using-deep-learning.html

https://www.mathworks.com/help/deeplearning/ug/multiple-input-and-multiple-output-networks.html

You can also use combineDatastore for your purpose.

https://www.mathworks.com/help/matlab/ref/matlab.io.datastore.combineddatastore.html

Alternatively, you can also create a custom layer that accepts multiple inputs.

https://www.mathworks.com/help/deeplearning/ug/define-custom-layer-with-multiple-inputs.html