How to train a deep neural network with 2 inputs

Hello, I just started to learn deep learning using Matlab. I had a problem with this two input model. Did I create the right model in Matlab? How to train the model (how to define the two input)?

Python:

class Deep_unfold_NN(torch.nn.Module ):
    def __init__(self,x_size, y_size):
        super(Deep_unfold_NN, self).__init__()
        self.s1 = torch.nn.Sequential(torch.nn.Linear(x_size, x_size))
        self.aux  = torch.nn.Sequential(torch.nn.Linear(y_size, x_size) )
       

    def forward(self,x,y,L):  
        
        for index in range(L):
            x=self.s1(x)+self.aux(y)
        return x
            
            

Matlab using deepNetworkDesigner ( L in the python code above is 4):

After the model is created, I exported it in Matlab and save the layers as layers1

        input={Input_1_train,Input_2_train};
        net_deepmimo = trainNetwork(input,Output,layers1,options);
        

The error information is :

Network: Invalid input layers. If the network has a sequence input layer, then it must not have any other input layers.

How should I build this model in Matlab?

Thank you very much.

Answer 1

[Srivardhan Gadila]

As per my knowledge and above information, I think that you can use imageInputLayer instead of sequenceInputLayer. To define and train a deep learning network with multiple inputs, specify the network architecture using a layerGraph object and train using the trainNetwork function by specifying the multiple inputs using a combinedDatastore or transformedDatastore object. Also you have to use the following syntax of trainNetwork

net = trainNetwork(ds,layers,options)

and not net = trainNetwork(X,Y,layers,options)

If you are unable to train using trainNetwork then you can create a custom training loop using automatic differentiation. To learn more, see Define Custom Training Loops.

You can refer to the following resources for more information: How can I train multi-input deep network without DataStore, Multi-Input CNN for image classification, Input Arguments & ds of trainNetwork and Multiple-Input and Multiple-Output Networks, Input Datastore for Training, Validation, and Inference