Multiple outputs in a deep sequential model

I would like to train a sequence model that predicts multiple binary outputs at each step of the sequence. However, when I try to design such a network, I get an error saying that networks with sequence input layers can have only one output (see attached figure).

Is there a solution to this problem?

Answer 1

[Shubham Rawat]

Hi Nathan,

You can only have a single output for a sequential input model. You can obtain your results by designing different models for different outputs, means copy your model each having different outputs and run all models and then get individual results.

You can also obtain multiple outputs if you don’t use sequential input. Here is the link below:

https://www.mathworks.com/help/deeplearning/ug/train-network-with-multiple-outputs.html

[Giorgio Grioli]

Hi Nathan and Shubham,

your question and relpy really helped me srot out a similar problem I was having in my work.

I am wondering if this limitation is due to some theorethical obstacle, or if it is only due a technical limitation of the current implementation. Would it be the second case, do you have any idea if Mathworks plan to implement support for this feature in the future?

I really hope you would be able to reply.

Best

Giorgio