MISO System. LSTM

Hello. I have a MIMO (multiple input-single output) system. Each of the two inputs has 100 random values. The only output has 100 values.

I put the entries as an array of cells:

sim_in = {v , p}';

and I pick up the output:

sim_out = {h1.Data}';

When training the network I get the error "Invalid training data. Predictors and responses must have the same number of observations." What would be the solution? Thank you.

Answer 1

[Aman]

Hi Jesus,

This error occurs due to a dimension mismatch between the targets and the network output.

Since you are using a MISO neural net, it is better to concatenate the inputs along the appropriate dimension and reshape your output data, as we give a single input and single output to the "trainNetwork" function.

You can refer to the thread below, which discusses the same issue.

https://in.mathworks.com/support/search.html/answers/2014196-invalid-training-data-predictors-and-responses-must-have-the-same-number-of-observations-in-trainn.html?fq%5B%5D=asset_type_name:answer&fq%5B%5D=category:deeplearning/applications&page=1

Hoping it will help you move ahead with your workflow.

[Jesus Mª Juarez Ferreras]

Thank you.

Answer 2

[Jesus Mª Juarez Ferreras]

Hello again. Once the previous problem is solved, now I put:

sim_out = {h1.Data h1.Data}';

I want to train the MISO system that I leave in the file. I want to get the RMSE close to zero but it doesn't train well.

Does LSTM not train MISO systems?

Thanks you.