On connectionism and the problem of correspondence