In particular, it has both global convergence and local superlinear convergence with less computation and weaker assumptions.

• 并且算法不再计算广义投影矩阵，使计算量大大减小。