Multi Input Multi Sequence Neural Network

The question is how to define a multi input multi sequence neural network (NN) in Matlab?

There is a way to define multi sequence NN:

 x1 =  [4 5 6];
 x2 =  [4 5 6];
 y1 = [1 2 3];
 y2 = [1 2 3];
 x = {[x1;x2]};

 y={[y1;y2]};
 net = fitnet(1);
 net = train(net,x,y);
 view(net)

 https://www.mathworks.com/help/nnet/ug/multiple-sequences-with-dynamic-neural-networks.html

There is also a way to define multi input NN:

 x1 =  [4 5 6];
 x2 =  [4 5 6];
 y1 = [1 2 3];
 y2 = [1 2 3];
 x = {x1;x2};
 y=y1;
 net = fitnet(1);
 net.numinputs = 2;
 net.inputConnect = [1 1; 0 0];
 net = train(net,x,y);
 view(net)

 https://www.mathworks.com/matlabcentral/answers/355286-how-to-give-multiple-inputs-to-the-train-function-of-neural-network

When I try to combine these two conditions:

 x1 =  [4 5 6];
 x2 =  [4 5 6];
 y1 = [1 2 3];
 y2 = [1 2 3];

 x = {{[x1;x2]};{[x1;x2]}};

 y={y1;y2};
 net = fitnet(1);
 net.numinputs = 2;
 net.inputConnect = [1 1; 0 0];
 net = train(net,x,y);
 view(net)

I get error:

 Error using nntraining.setup>setupPerWorker (line 61)
 Inputs X{1,1} is not numeric or logical.

 Error in nntraining.setup (line 43)
 [net,data,tr,err] = setupPerWorker(net,trainFcn,X,Xi,Ai,T,EW,enableConfigure);

 Error in network/train (line 335)
 [net,data,tr,err] = nntraining.setup(net,net.trainFcn,X,Xi,Ai,T,EW,enableConfigure,isComposite);

Answer 1

[Kenta]

As of now, you can use custom training loop for deep learning for the multi-input. This is a bit different from what you want, but I believe it helps you.

https://jp.mathworks.com/matlabcentral/fileexchange/74760-image-classification-using-cnn-with-multi-input-cnn