Signaldatastore for sequence to one deep learning regression

Hi, I have a large matrix A comprising time series of size 4000 x 20 000 with the first dimension being the number of predictors and the second the number of samples.

I have a vector of size 1 x 4000 comprising one value per time series.

I would like to train a LSTM network akin to the tutorial on sequence-to-one deep learning regression.

I split A into training and test subsets and store each time series as a .mat file in either a test_data folder or a train_data folder.

I would like to make two signal data stores, ds_train and ds_test referencing the test and and train data respecitively. I successfully do this. However, I cannot seem to also reference the labels - how do I do that? Do I need to also make a separate .mat file for each label corresponding to each time series in each of the test_data folder and train_data folder?

Any help is greatly appreciated.

Kind regards

Knut Skaug

[Debraj Maji]

Hello @Knut Skaug,

It would be great if you could provide the code for the network?