Well, I did a project in NN when I was in my college. As much I remember, I think training a NN is a part of completing a process. Training here means providing certain sets of inputs and getting back desired output. And for getting desired output, you need to have certain weights.
I found a good link that should explain you what and how you can train your NN, look
here[
^].
If you get stuck somewhere related to MATLAB, then visit
here[
^].