[rottjung]

ok so this is how i revert the result to my original scale, thanks for the help guys:
Mathf.RoundToInt(sig.InputScaler((-Mathf.Log((1-results[0])/results[0])),-4f,4f,0f,10f)
works perfect.

It's starts to look promising :-) still some hiccups because of me sucking at math tho.
I'll move progress to the members experience section.

hey spydaz i want to use the NN to approximate a function, for example: 1+1=2. my input would be 1 my desired result for training 2.
if the NN is trained it adjusted it's weights to output 2 with other words the NN function approximates +1
putting 3 as input to the trained NN should result in 4 and so on.

so what mine should do is approximates the change of a value's x,y coordinates from matrix1 in comparison to the value's x,y coordinates in matrix2. this way it should have learned the causes of change from matrix1 to matrix2.