Feed forward Neural network (back propagation) code classification problem

[C# Programming]

Hi, Everbody.

i have writen c# Artifical intelliget Neural Network. I have 500 count dataset which is normalized -1 and 1 and.i have used learning rate and momentum coefficient .

i teaching my network with 400 count dataset. Network have 5 inputs and 5 outputs. my output is setuped as 00001/00010/00100/01000/10000
when my network teched with 10000 iterations and 400 datasets, my output have a problem.
my network teached in last class of 10000 so when i tested my datasets , output only have been predicted as true class of 10000.
as result my code can only learn last class and other classes can not predicted or poor predicected

Example Test and Learning dataset
5 neuron input

{{0.04418 ,0.18135 ,0.28143 ,0.34367 ,0.14938 ,},
{0.37418 ,0.35050 ,0.18184 ,0.07493 ,0.01855 ,},
{0.04540 ,0.17061 ,0.24506 ,0.32560 ,0.21333 ,},
{0.08494 ,0.25653 ,0.26922 ,0.26141 ,0.12790 ,},
{0.18135 ,0.36832 ,0.24799 ,0.15743 ,0.04491 ,},
{0.11887 ,0.29143 ,0.27581 ,0.22870 ,0.08518 ,},
{0.02758 ,0.26239 ,0.37247 ,0.29998 ,0.03759 ,},
{0.00781 ,0.15792 ,0.32145 ,0.41811 ,0.09470 ,},
{0.07811 ,0.31608 ,0.31901 ,0.24506 ,0.04174 ,},
{0.07664 ,0.32390 ,0.32829 ,0.21674 ,0.05443 ,},
....


5 neuron output
{
{1 ,0 ,0 ,0 ,0 },
{1 ,0 ,0 ,0 ,0 },
{0 ,1 ,0 ,0 ,0 },
{0 ,1 ,0 ,0 ,0 },
{0 ,1 ,0 ,0 ,0 },
{0 ,0 ,1 ,0 ,0 },
{0 ,0 ,1 ,0 ,0 },
{0 ,0 ,0 ,1 ,0 },
{0 ,0 ,0 ,1 ,0 },
{0 ,0 ,0 ,0 ,1 },
{0 ,0 ,0 ,0 ,1 }
.....


My trainin code blok is as folowing show


int iterasyon=100000;
while (it = 0.00001)//iteration start section
{
mmse = 0;
mae = 0;
for (int nf = 0; nf <span class="code-keyword">