Global convergence properties of the two new dependent Fletcher–Reeves conjugate gradient methods☆
Introduction
Consider the unconstrained optimization problemwhere f : Rn → R is continuously differentiable. Conjugate gradient methods for solving (1.1) are iterative methods of the formwhere gk = ∇f(xk), αk is a step-length obtained by a one-dimensional line search and βk is a scalar. The well-known formulas for βk are the Fletcher–Reeves (FR) [18] formulaand the Polak–Ribière–Polyak (PRP) [19], [20] formulawhere ∥ · ∥ means the Euclidean norm. The conjugate gradient method is available for large-scale unconstrained optimization because its storage is relatively small. Number results in [17] shows that if f is easy to be computed and if its dimension n is very large, the conjugate gradient method is still the best choice for solving (1.1). Convergence properties of the FR conjugate gradient method have been studied by many authors, including [1], [3], [4], [5], [6], [7], [8], [9], [10], [13], [15], [16], [18], [19], [20].
Many efforts have been devoted to investigate the global convergence properties of the FR conjugate gradient method. Powell [8] and Zoutendijk [13] proved that the FR conjugate gradient method with exact line searches is globally convergent on general functions. Al-Baali [1] extended this result to inexact line searches. However, the numerical performance of the FR conjugate gradient method is often much slower than that of the PRP conjugate gradient method. The PRP conjugate gradient method has a good numerical performance [9], but does not have such good convergence property [8]. Due to this, some dependent FR conjugate gradient methods are proposed [5], [6], [7], [15], [16]. In this kind of methods, the region of βk with respect to is introduced and if βk is chosen suitably in this region, then the method not only has the global convergence property but also has the nice numerical experience.
Gilbert and Nocedal [5] investigated global convergence properties of the dependent FR conjugate gradient method with βk satisfying , provided that the line search satisfies the strong Wolfe conditionswhere 0 < δ < σ < 1. If βk satisfies (where c > 1), they have given an example to indicate that even exact line search cannot guarantee the global convergence property of the dependent FR conjugate gradient method.
Hu and Storey [6] proved the dependent FR conjugate gradient method is globally convergent with the strong Wolfe line search if βk satisfieswhere 0 < σ < 1, , and c > 0. Liu et al. [7], [16] had the same result if . It should be pointed out that the global convergence of the previous dependent FR conjugate gradient method can be obtained only under the strong Wolfe line search.
Dai and Yuan [4] investigated global convergence properties of the FR conjugate gradient method with the generalized Wolfe conditions, i.e. (1.6) andwhere 0 < δ < σ1 < 1 and σ2 > 0. They also proved that σ1 + σ2 ⩽ 1 is a sufficient condition for ensuring for all k. It is easy to see that (1.7) can be viewed as a special case of (1.9) with σ1 = σ2.
Dai [3] also investigated global convergence properties of the FR conjugate gradient method with the Armijo line search, which is to determine the smallest integer m ⩾ 0 such that, if one defines αk = λm, thenwhere λ ∈ (0, 1) and δ ∈ (0, 1).
Wang and Zhang [15] proposed an s-dependent GFR conjugate gradient method and made two different kinds of estimations of upper bounds of βk with respect to where , , s ∈ {1, 2}; k ⩾ 3, Ω1 is a partition of set {1, 2, … , t}, t ⩾ 1 is an integer and the definition of is as follows:They proved global convergence of the s-dependent GFR conjugate gradient method with some various line searches.
In this paper, we give new restriction to βk which is wider than the previous restriction, i.e. βk satisfieswhere c > 0 is a constant, t ⩾ 1 is an integer and s ∈ {1, 2}. Further, we also define when Ω = ϕ.
Generally, the two regions of βk defined by (1.14) do not conclude each other.
In Section 2, we will investigate global convergence properties of the two new dependent FR conjugate gradient methods with the generalized Wolfe line search (1.6), (1.9), the Armijo line search (1.10) and the Wolfe line search respectively. In Section 3, we will indicate that some of the global convergent results in [3], [4], [5], [6], [7], [15], [16] can be viewed as a special case of the global convergent results in Section 2.
Section snippets
Main results
In this section, we always assume that ∥gk∥ ≠ 0 for all k, for otherwise a stationary point has already been found. We make the following assumption on the objective function. Assumption 2.1 f is bounded below on the level set . The level set is bounded. In some neighborhood of , f is differentiable and its gradient g is Lipschitz continuous, namely, there exists a constant L > 0 such that
We state a general convergence result as follows. This
Some remarks
From the results in Section 2, we can deduce some known results. First, letting c = 1 in (1.14), we know that βk satisfies (1.14) when it satisfies . Thus Theorem 2.3 in [4] and Theorem 3.2 in [5] are the special cases of Theorem 2.1, and Theorem 1 in [3] is a corollary of Theorem 2.3.
Second, we assume that βk satisfieswhere 0 < σ < 1, , , C > 0. Eqs. (3.1), (3.2) can be used instead of (1.8) under the assumption (B). With the strong Wolfe line
Numerical results
In this section, we report numerical results for the two dependent FR conjugate gradient methods. For comparison purposes, we also coded the FR conjugate gradient method. Eight test problems having the form of are selected from Ref. [21]. Numerical results are given in Table 1, Table 2 (where m is the number of function fi and n is the dimension of x). In Table 1, Table 2, “Iter” denotes the number of iteration, “NF” denotes the number of function evaluation, and “NG” denotes
References (21)
The conjugate gradient method in extremem problems
USSR Comput. Math. Math. Phys.
(1969)Descent property and global convergence of the Fletcher–Reeves method with inexact line search
IMA J. Numer. Anal.
(1985)- et al.
A tool for the analysis of quasi-Newton methods with application to unconstrained minimization
SIAM J. Numer. Anal.
(1989) Further insight into the convergence of the Fletcher–Reeves method
Sci. China
(1999)- et al.
Convergence properties of the Fletcher–Reeves method
IMA J. Numer. Anal.
(1996) - et al.
Global convergence properties of conjugate gradient methods for optimization
SIAM. J. Optim.
(1992) - et al.
Global convergence result for conjugate gradient methods
JOTA
(1991) - et al.
Global convergence of the Fletcher–Reeves algorithm with inexact line search
Appl. Math. Chin. Univ., Series B
(1995) Non-convex minimization calculation and the conjugate gradient method
(1984)Restart procedure for the conjugate gradient method
Math. Prog.
(1977)
Cited by (1)
Inverse Dynamic Analysis of an Inclined FGM Beam Due to Moving Load for Estimating the Mass of Moving Load Based on a CGM
2020, Iranian Journal of Science and Technology - Transactions of Mechanical Engineering
- ☆
This work is supported by N.S.F. of China (10571106) and the foundation of Qufu Normal University.