Next: DLVQ:
Up: Special Network Architectures
Previous: TDNN:
Counterpropagation doesn't need the output layer. The
output is calculated as a weighted sum of the activations of the
hidden units. Because only one hidden unit has the activation 1 and
all others the activation 0, the output can be calculated with the
winner unit, using the weights from this unit to the output.
Niels Mache
Wed May 17 11:23:58 MET DST 1995