Hi Robin Patel,
It is my understanding that you are trying to build a custom Neural Network for multiple output regression.
As per the Neural Network given in attached image, the network can be built by creating a graph of Network layers using "lgraph" function and then creating a "dlnetwork" object which can be trained using custom training loops.
Refer to following code to create a Neural Network according to the attached image.
Create a main branch of the Network.
layers = [featureInputLayer(17,'Name','in') % Input layer for 17 features
fullyConnectedLayer(10) % fully connected layers
tanhLayer('Name','tanh')
fullyConnectedLayer(10,'Name','fc1')
tanhLayer("Name",'tanh1')
fullyConnectedLayer(1,'Name','Out1')
];
Convert it into Layer Graph
lgraph = layerGraph(layers);
Add layers of the second branch to the Layer Graph
lgraph = addLayers(lgraph,[fullyConnectedLayer(10,'Name','fc2'), ...
tanhLayer("Name",'tanh2'), ...
fullyConnectedLayer(1,'Name','Out2')]);
Connect the second branch to the main branch. Then convert the whole Layer Graph into "dlnetwork" object.
lgraph = connectLayers(lgraph,'tanh','fc2');
dlnet = dlnetwork(lgraph);
figure
plot(lgraph)
Refer to the following documentation link to learn more about "layerGraph".
Refer to the following documentation link to learn more about "dlnetwork" object.
Refer to the example in the following documentation link to learn more about Multiple Output Neural Network.
Hope this helps!
