Enhancing convergence toward consensus in leader-follower networks