Can Humans Do Less-Than-One-Shot Learning?