针对非线性等式和不等式约束优化问题,结合罚函数法,提出了一个共轭投影梯度法,在适当的假设条件下,证明了该方法的全局收敛性,由于使用了目标函数的二阶导数信息,该算法的超线性收敛速度也能得到保证,数值结果表明,该算法是有效的。
A conjugate gradient projection method for nonlinear equality and inequality constrained optimization problems was proposed combing with the method of penalty function and its global convergence proved under suitable assumptions. Superlinear convergence of the iteration could also be guaranteed because of using second derivatives of function message. Some numerical experiment results showed that the given method is efficient.