【Numberical Optimization】5 Conjugate Gradient Methods (zen学习笔记)

5.1  The linear conjugate gradient methods

The conjugate gradient method is an itterative method for solving a linear system of equations:

                                                                                            【Numberical Optimization】5 Conjugate Gradient Methods (zen学习笔记)

which can be stated equivalently as the following minimization problem:

                                                                            【Numberical Optimization】5 Conjugate Gradient Methods (zen学习笔记)

第k步的残差:【Numberical Optimization】5 Conjugate Gradient Methods (zen学习笔记)

5.1.1  Conjugate direction methods

1.阐述了共轭方向算法能在有限步内收敛到最优点(步数主要与特征值分布有关)

The conjugate vectors are given.

  • 【Numberical Optimization】5 Conjugate Gradient Methods (zen学习笔记) is said to be conjugate with respect to the symmetric positive definite martix A if 【Numberical Optimization】5 Conjugate Gradient Methods (zen学习笔记)
  • 【Numberical Optimization】5 Conjugate Gradient Methods (zen学习笔记) is also linearly independent.
  • 有限步内能收敛到最优解,证明及算法如下:

 【Numberical Optimization】5 Conjugate Gradient Methods (zen学习笔记)

Defining 【Numberical Optimization】5 Conjugate Gradient Methods (zen学习笔记) as 

                                                                                      【Numberical Optimization】5 Conjugate Gradient Methods (zen学习笔记)

where S is defined by 【Numberical Optimization】5 Conjugate Gradient Methods (zen学习笔记)

then 【Numberical Optimization】5 Conjugate Gradient Methods (zen学习笔记)

2.最优点在集合中      【Numberical Optimization】5 Conjugate Gradient Methods (zen学习笔记)即可由【Numberical Optimization】5 Conjugate Gradient Methods (zen学习笔记)线性表出

【Numberical Optimization】5 Conjugate Gradient Methods (zen学习笔记)

5.1.2  Basic properties of the conjugate gradient method

          A PRACTICAL FORM OF THE CONJUGATE GRADIENT METHOD

前面介绍了共轭梯度算法的一些性质,接下来的算法提出共轭方向【Numberical Optimization】5 Conjugate Gradient Methods (zen学习笔记)的构造

共轭梯度初步算法及共轭梯度算法

【Numberical Optimization】5 Conjugate Gradient Methods (zen学习笔记)【Numberical Optimization】5 Conjugate Gradient Methods (zen学习笔记)

改进之处在于:

  • ​​​​​​​【Numberical Optimization】5 Conjugate Gradient Methods (zen学习笔记)
  • 【Numberical Optimization】5 Conjugate Gradient Methods (zen学习笔记)

5.1.3  RATE OF CONVERGENCE

解释收敛步数与特征值分布的关系、验证 CG算法在某种意义上是最优的