Mountain View, CAUSA 94043
September 24, 2010
Hi, this is Pratibha. I love playing, watching and discussing tennis :-)
Sat evening / Sunday
No answer yet
Hi Pratibha! Welcome to Velo Girls. I look forward to riding with you soon.
Heya, I just thought about the whole vectorizing thing a bit more...
It doesn't seem like it's possible to "vectorize" all three layers at once, since calculating all the layers can't be represented by matrix multiplications; you can, however, abstract the backpropagation method into a function, whose signature would be something like:
backpropagate(int layers, int activation, int y, inttheta1, inttheta2, ...)
The function would then go through each layer.