Why cannot Matlab train multiple time series data directly?

I don't see the major technical difficulty here. But it seems that Matlab doesn't provide such a capability. For training a model, I would like the model to predict multiple time series data with a good accuracy. So, MSE calcualtion should consider all avaiable time series data. Training individually and update one after the other is not good option. Also, connect multiple time series data to form one time series data is also not a good option. Any body knows the way to do it in Matlab?

[Walter Roberson]

Do I understand correctly that you have several different timeseries that are each measuring a different variable, and that the variables interact so you need information from all of them together to make predictions? And do I also understand correctly that the time series have different associated times, and that is why you cannot merge them into a single timeseries without undersired interpolation?

[Zhimin Xi]

Y(t) = f(x1(t),x2(t),...,xn(t)), (e.g., t = 1:100). That forms a basic model and we don't know f( ) which I wanna use a time-delayed RNN to identify the f( ).

Assume that I can collect 10 samples (or observation) for the model and each sample means data observations for Y(t) = f(x1(t),x2(t),...,xn(t)), (e.g., t = 1:100). Now the true model form cannot be found simply based on each sample observation (let's assume I know this as a fact), and connection of all 10 samples will make undesirable connections espeically with the time delayed effects between samples.

[Walter Roberson]

How does that differ from using a sequence as input?

https://www.mathworks.com/help/deeplearning/ref/nnet.cnn.layer.sequenceinputlayer.html