【摘要】 对无约束规划问题,本文提出了结合Armijo步长搜索规则的一类带误差项的记忆梯度求解算法,并在目标函数的梯度一致连续的条件下,证明了算法的全局收敛性。同时给出带误差项的结合拟-Newton方程的记忆梯度算法。数值例子表明算法是有效的。更多还原

【Abstract】 We consider the convergence properties of a new memory gradient method with errors and Armijo step size rule for unconstrained optimization problem,under the assumption that the gradient of the function is uniformly continuous.Combining the quasi-Newton equation with our new method, quasi-Newton method with errors is modified to achieve the global convergence property.Numerical results show that the new algorithms are efficient.更多还原