Neural Network -- Problems with stopping condition

My source is here: Download Project_2_redo.rar from Sendspace.com - send big files the easy way
I believe my problem is my stopping condition. For example my code is supposed to receive input from a 14 digit led and return the corresponding number My problems is that it seems to forget what it learned. Also it seems to go way to fast so i believe something is wrong. for example there are different ways to write numbers on a 14 digit display so my code learns the first and it takes a few iterations then it learns the next version of that number swiftly. then it takes a while to learn the next number. My stopping condition is to keep chaining the weights until the final user readable outputs are the same as the desired outputs, once they match it has learned that number. When it "learns" all the numbers I give it another number to see if it understands it and if just gives me the last 'learned' number. I don't know what I am doing wrong, can someone please help me.

my gmail is: [ deleted, to protect the OP from loads of viagra and very rich old women from Nigeria etc. - Jos ]