一类无约束优化混合共轭梯度法的全局收敛性  

Global Convergence of A hybrid Conjugate Gradient Method for Unconstrained Optimization

作　　者：翁世有[1] WENG Shi-you(Mathematics Department of Suzhou Vocational University,Suzhou 215104,China)

机构地区：[1]苏州市职业大学数理部,江苏苏州215104

出　　处：《辽宁高职学报》2020年第5期80-83,共4页Journal of Liaoning Higher Vocational

摘　　要：共轭梯度法是求解无约束问题的一类有效方法。提出一类新的共轭梯度法,该方法在wolfe搜索条件下,保证了算法的全局收敛性。其优点是,保留允许参数取负值,并改善了PRP方法收敛性不好和DY不一定产生下降方向的问题,具有收敛性好,收敛速度快的特点。从数值结果中表明该算法可行、有效。Conjugate gradient method is an effective method for solving unconstrained problems. A new conjugate gradient method is proposed, which guarantees the global convergence of the algorithm under wolfe search conditions.The advantages of this method are that the negative value of parameters is allowed, and the bad convergence of PRP method and the problem that DY does not necessarily lead to falling direction are improved, so it has the characteristics of good convergence and fast convergence. The numerical results show that this algorithm is feasible and effective.

关 键 词：无约束优化 共轭梯度法 线搜索 全局收敛 

分 类 号：O224[理学—运筹学与控制论]