We turn next to the task of finding a weight vector w which minimizes the chosen function E(w). Because there is clearly no hope of finding an anlytical solution to the equation ∂E(w)=0, we resort to iterative numerical procedures. On-line gradient d