Is backwards pass implemented correctly?

[a3cel2]

I've noticed there's a factor of 100 division in the forward pass, but no similar factor in the backwards pass. I also noticed that setting percentiles as a parameter in my network does not make these parameters change much from their initial value over many epochs, even though all of the other variables in the network seem to. Removing the division with 100 in the forward pass seems to fix it (such that input is given between 0-1), but I'm not sure if I'm doing something wrong by doing this