LSTM neural network for regression

Hello, I would like to create a recurrent neural network for a regression problem. I'd like to use a LSTM network with a continuous output, but I'm getting an error when building the network with a LSTM Layer and Regression Layer: "Regression is not supported for networks with LSTM layers."

Is there a way I can work around this to have a simple, continuous output from -1 to 1 instead of just a classification?

My input is [720 10] - there are 720 features and 10 samples over time. The feature values are roughly between -10 and 30. My output should be [12 1] - there are twelve output variables and each has a value ranging from -1 to 1.

Thanks, Jake

[Greg Heath]

For N pairs of I-dimensional inputs and corresponding O-dimensional "O"utput targets, the matrix dimensiona are

 [ I N ] = size(input)

 [O N ] = size(target)

 with, typically, 

 N > 10*max(I,O)

It is not clear to me what your values of I, O and N are

[Jake]

The time sequence data is continuous. The input features are adjustable. The length of a given sequence is adjustable as well.

The question is more, how do I get MATLAB neural network with an LSTM layer to work with a regression layer...

[Michael Henderson]

Can an LSTM layer be used with a regression layer?