Yes thanks. Much more clear that way. Edit: The talk was given by a geeky grad student, so perhaps very forward-thinking. Was quite a departure from the usual topics like Perceptron, MiniMax, Bayesian techniques...