Backpropagation in equations III

• The backpropagation learning rule (applied at time t) is:

  ?wji(t) = ? ?jai + ??wji(t-1)
  

• where ?wji (t) is the change in the weight from node i to node j at time t,

• The learning constant ? is typically chosen rather small (e.g., 0.05).

• The momentum term ? is typically chosen around 0.5.