Multiple output convolutional nueral netwok

Hello, I have been using the DAGNetwork functionality to put together a netwok that requries multiple outputs, the plan is to do some multi task learning, I tried to go for a multi loss function that is optimized jointly then feeds to one output but that proved difficult with Matlab since most of this stuff isn't supported and I haven't been able to get any help here. now I'm looking into optimizing them seperately using multiple outputs after sharing layers.

Is any of this possible with matlab?

[Alla]

The mroe I look into it the clearer it is that these things are not possible on Matlab, too bad, matlab's deep learning toolbox is very useful, easy to use and well documented but it seems that it doesn't allow for much beyond the standard deep learning functionalities.