Multi-output network design in "Deep Network Designer"

Hi,

Is it possible to design single input/multi output network in Deep Network Designer"? If so, is there any example to show this case.

[KSSV]

What is your problem? Wha for you are targetting Deep Learning?

[Samen P]

I am using 2d spectrograms and have 10 outputs where each one could be 0 and 1. It is possible that more than one output be 1.

Answer 1

[Yomna Genina]

It is possible to design a multi-output network in Deep Network Designer, however it is not possible to train it using Deep Network Designer as training multi-output networks is not support with trainNetwork.

To train a multi-output network, you could build the network first using Deep Network Designer then export the network to workspace, then convert the layer graph to a dlnetwork and use custom training to train it.

Here is an example on creating a multi-output layer graph, converting the layer graph to dlnetwork then using custom training to train the multi-output network:

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