Backpropagation: Recombine the Error

Recombination of errors

The training data examples only tell us what the outputs from the final nodes should be. They don’t tell us what the outputs from nodes in any other layer should be. This is the core of the puzzle.

We could recombine the split errors for the links using the error backpropagation we just saw earlier.

So, the error in the first hidden node is the sum of the split errors in all the links connecting forward from the same node. In the diagram above, we have a fraction of the output error $e_\text{output,1}$ on the link with weight $w_{}$ ...