Hi, I'm having some trouble finding out exactly what the widrow-hoff learning rule is. In the notes I have, they say that its: NewWeight = OldWeight + LearningRate*(DesiredOutput - OldWeight*Input) It even comments that this rule is completely independent of the transfer function, but I've seen so many different versions of this rule that I dont know which to believe. I've seen it written that the widrow-hoff rule is the same as the delta rule, but in my notes they're completely different. In another set of notes I have the same rule listed as: NewWeight = OldWeight + LearningRate*(DesiredOutput - ActualOutput)*(Input/(length of inputVector)^2) So can anyone tell me what the right rule is? Thanks.

If you want the rule as originally used by Widrow and Hoff, I suggest returning to their original paper on 'Adaline'.