How do I create a neural network that will give multiple outputs?

I have data in the form 141x18, i.e., there are 141 cases with 18 parameters each. I want to create a feedforward network that is capable of producing a weight matrix of 18x18, which is the weights from the second hidden layer to the output layer. I am only able to produce an output layer of 141x1. I am able to produce a network with structure 18-36-36-1, however, is it possible to create a network of 18-36-36-18? If change the net.outputConnect(2) = 18; I get the error: >Error using network/train (line 340) >Number of targets does not match net.numOutputs.

Please help. My script is :

      clear;
      net= feedforwardnet([36 36]);
      net.Inputs{1}.size = 18; %(Number of neurons in the input layer)
      %net
      numLayers = 2; %(Total numbers of layers)
      net.Layers{1}.size = 36;%(Number of neurons in first layer)
      net.Layers{2}.size = 36; %(Number of neurons in first layer)
      net.inputConnect(1) = 1;
      net.layerConnect(2,1) = 18;

      net.outputConnect(2) = 18;
      net.targetConnect(2) = 1;
      net.layers{1}.transferFcn = 'tansig';
      net.layers{2}.transferFcn = 'purelin';
      %net.biasConnect = [1;1];
      net = init(net);
      net.initFcn = 'initlay';
      net.layers{1}.initFcn = 'initnw';
      net.layers{2}.initFcn = 'initnw';
      net.inputWeights{1,1}.initFcn = 'rands';
      net.inputWeights{2,1}.initFcn = 'rands';
      net.biases{1}.initFcn = 'rands';
      net.biases{2}.initFcn = 'rands';
      net.performFcn = 'mse';
      net.trainFcn = 'trainlm';
      net.trainParam.lr = 0.02;
      net.trainParam.goal = 0.00010;
      net.trainParam.mc = 0.09;
      net.trainParam.epochs = 10000;
      net.trainParam.show = 100;
      numNN = 10;
      nets = cell(1,numNN);
      %net.input.processFcns = {'removeconstantrows','mapminmax'};
      %net.output.processFcns = {'removeconstantrows','mapminmax'};

      % Setup Division of Data for Training, Validation, Testing
      % For a list of all data division functions type: help nndivide
      net.divideFcn = 'dividerand';  % Divide data randomly
      net.divideMode = 'sample';  % Divide up every sample
      net.divideParam.trainRatio = 90/100;
      net.divideParam.valRatio = 5/100;
      net.divideParam.testRatio = 5/100;

      % Choose a Performance Function
      % For a list of all performance functions type: help nnperformance
      net.performFcn = 'mse';  % Mean Squared Error

      % Choose Plot Functions
      % For a list of all plot functions type: help nnplot
      net.plotFcns = {'plotperform','plottrainstate','ploterrhist', ...
          'plotregression', 'plotfit'};

      load data.mat

      A = transpose(A1);
      B = transpose (B1);
      [An,minA,maxA] = premnmx(A)
      [Bn,minB,maxB] = premnmx(B)
      net = init(net);

      [net,tr] = train(net,An,Bn);
      %gensim (net);
      a1 = sim(net,An);
      a = postmnmx(a1,minB,maxB);
      t = B - a;
      perf = mse(t)
      net.performFcn = 'mse';
      x=[1:1:141]  
      plot(x,a,'r+:',x,B,'gd:')
      % Test the Network
      y = net(A);
      e = gsubtract(B,y);
      performance = perform(net,B,y)

      % Recalculate Training, Validation and Test Performance
      trainTargets = B .* tr.trainMask{1};
      valTargets = B .* tr.valMask{1};
      testTargets = B .* tr.testMask{1};
      trainPerformance = perform(net,trainTargets,y)
      valPerformance = perform(net,valTargets,y)
      testPerformance = perform(net,testTargets,y)

      IW = net.IW{1,1};
      LW = net.LW{2,1};
      bias = net.b{1};
      IW = net.IW{1,1};
      LW = net.LW{2,1};
      bias = net.b{1};

Answer 1

[Greg Heath]

x = randn(18,141);
t = randn(18,141);
net = feedforwardnet([ 36 36 ]);
net = train(net,x,t);
view(net)

Thank you for formally accepting my answer

Greg

[Pkm]

How did you solve this? net.outputConnect is 1*N matrix, where N is number of layers. Whereas your output is 18? I'm facing same problem too! @Greg I tried to give the same way as you have given but it doesn't work as my data is 1x600 cell array of 960x1 matrices, with 600 timesteps of 960 elements.