input size error in neural network programming by MATLAB

Question

This is my code:

p = input1;
t1 = output1;
net = feedforwardnet(10, 'trainrp');
net.trainParam.epochs = 1000;
net.trainParam.goal = 0.0005;
net = train(net, p, t1); 
y1 = sim(net, p);
p = input2;
t2 = tar;
y2 = sim(net, p);

However, I get this error:

error using bsxfun
Non-singleton dimensions of the two input arrays must match each other.

Error in nnMATLAB.pc (line 24)
pi = bsxfun(@minus,pi,settings.xoffset);

Error in nncalc.preCalcData (line 20)
data.Pc = calcMode.pc(net,data.X,data.Xi,data.Q,data.TS,calcHints);

Error in nncalc.setup1 (line 118)
calcData =

nncalc.preCalcData(matlabMode,matlabHints,net,data,doPc,doPd,calcHints.doFlattenTime);

Error in network/sim (line 283)
[calcMode,calcNet,calcData,calcHints,~,resourceText] = nncalc.setup1(calcMode,net,data);

I want to make a neural network which has input1 as input 310 x 24 matrix and output1 as output, 155 x 24 matrix.

Also, I will train the network with input1 & output1

After this training process, I will use input2 as testing data, and I want to get a simulation result using above network which is trained by input1 & output1 .

In summary, I want to train my own network with input1 and output1, and I want to get my simulation result with input2.

I think these errors are based on input size difference between training section and testing section.

How can I solve this problem? Should I run additional process??

I'm looking for your kind answer.

Thank you.

Mark Hudson Beale Mark Hudson Beale · Accepted Answer · 2014-03-26T02:44:53

The number of inputs for training and testing must be the same.

For training: inputs must be an NxQ matrix where N is the number of input elements and M is the number of samples. Target must be an MxQ matrix where M is the number of output elements and Q is the same as for inputs.

Then for testing: the input matrix must be NxQ2, where N is the same as for training, but the number of samples Q2 can be whatever you want. For instance, for a single vector Q2 equals 1. The output will then be MxQ2 where M is the same as the number of outputs used for training and Q2 is the same number of vectors as the test input data.

input size error in neural network programming by MATLAB

1 Answers