Uniform Convergence of Rank-Weighted Learning