next up previous
Next: Backward Propagation Up: Back-Propagation algorithm Previous: Back-Propagation algorithm

Forward Propagation

The idea of the forward phase is shown in figure 7. The weights of the needed receptive connections of neuron j are one row of the weight matrix. The following values are calculated:
activation of neuron j : aj := ioi.wji+ &thetas;j

output of neuron j : oj := f(aj) = {11 + exp(-aj)}

Fri Jun 30 13:29:58 MET DST 1995