Train Recgnition Neural Network on image moements

I am trying to train pattern recognition neural network on set of image moments values, when trying to simulate i keep getting the same result for every input, tried to normalize the data using "mapminmax" but still no luck

P.S the seven moments contain negative numbers

[Greg Heath]

Insufficient explanation. Include relevant code and error message.

[Tareq Al-Zubaidi]

I am using matlab neural network wizard, the code generated :

inputs = ann_data_match_only_mapminmax';
targets = ann_data_match_only_mapminmax_truth_binry';

% Create a Pattern Recognition Network
hiddenLayerSize = 10;
net = patternnet(hiddenLayerSize);

% Setup Division of Data for Training, Validation, Testing
net.divideParam.trainRatio = 70/100;
net.divideParam.valRatio = 15/100;
net.divideParam.testRatio = 15/100;

% Train the Network
[net,tr] = train(net,inputs,targets);

% Test the Network
outputs = net(inputs);
errors = gsubtract(targets,outputs);
performance = perform(net,targets,outputs)

% View the Network
view(net)

when trying to simulate the result network on one entry of the training data

I = ann_data_match_only_mapminmax(1,:);
I = I';
I =

    1.0000
   -0.7889
   -0.7029
   -0.9304
   -0.9990
   -1.0000
   -0.9988
T = sim(net,I);
T = round(T);

>> T'

ans =

1 1 1 0 0 1 0

where i expect the output to be

1 0 1 1 0 1 0

I was think it is because of input data not normalized,, any idea why ?

Answer 1

[Greg Heath]

Mapminmax is a default.

What did you get for performance?

You may have converged to a local min.

Repeat 9 more times with different random initial weights.