Hi I want to implement multiple hidden layers for a BPN and am unsure as to how to back prop the error through more than one hidden layer. I have a single hidden layer BPN working fine, I am just unclear as to how to make the next step for multiple hidden layers. Any help or links would be much appreciated. Ta M

see the links page at my website

You don't need to use more than one hidden layer though, a single hidden layer can approximate any input/output mapping.

Quote: Original post by fup
see the links page at my website

[wow] ... blue links on blue background hurt me eyes!! You can define the colors for links in the body tag; I'm willing to bet that you're browser doesn't display them like that! (something like yellow would be nicer)

But yeah; I'd recommend Generation5 for AI stuff. And can't you backpropagate through multiple hidden layers? Maybe I'm not reading closely enough... [wink]

"blue links on blue background hurt me eyes!! You can define the colors for links in the body tag; I'm willing to bet that you're browser doesn't display them like that! (something like yellow would be nicer)
"

point taken. I'll try to rectify the color of the links later today.

"And can't you backpropagate through multiple hidden layers?"

Yes, you can propogate the error through as many hidden layers as you like.