It does seem likely that gates was misunderstood. > I suspect but cannot prove I think the rates of untrainable weights in a NN is proof enough, ie., that oft

It does seem likely that gates was misunderstood. > I suspect but cannot prove I think the rates of untrainable weights in a NN is proof enough, ie., that often around 30pc aren't being updated.