I am making simple leaf recognizing prorgam. I have 10 plant leaf data and total sample size about 660.
My input size 3, output layer 10. Hidden layers is changeable.(2 between 30)
First input data: compactness = (2*pi*Leaf Area) / (perimeter*perimeter)
Second input data: aspect Ratio = leaf windth / leaf height
Third input data: fullness = leaf Area / ( width * height )
And I am normalizing [0,1] all of data.
Output data is [1,0,0,0,0,0,0,0,0,0] for first plant species
[0,1,0,0,0,0,0,0,0,0] for second plant species
[0,0,1,0,0,0,0,0,0,0] for third plant species
[0,0,0,0,0,0,0,0,0,1] for tenth plant species.
But my error is about 32 and no recognize plant species.
My transfer function BipolarSigmoid and alfa value is 2. Learning Rate : 0.5 Momentum : 0.0
network = new ActivationNetwork(
new BipolarSigmoidFunction(2), 3, trackBar2.Value, 10 );
BackPropagationLearning backprob = new BackPropagationLearning( network );
What is my wrong? Thanks in advance. (sorry my english)