Mathematical and Computer Modelling 50 (2009) 15–20
Contents lists available at ScienceDirect
Mathematical and Computer Modelling
journal homepage: www.elsevier.com/locate/mcm
A PRP type method for systems of monotone equations
I
Wanyou Cheng
College of Computer, Dongguan University of Technology, Dongguan 523000, China
a r t i c l e i n f o
Article history:
Received 20 March 2007
Received in revised form 11 March 2009
Accepted 1 April 2009
Keywords:
Monotone equations
Hyperplane projection method
Global convergence
a b s t r a c t
In this paper, we propose an algorithm for solving systems of monotone equations. The
method is a combination of the well-known PRP method and the hyperplane projection
method. We prove that the proposed method is globally convergent if the equation is
monotone and Lipschitz continuous without any differentiability requirement on the
equation. Preliminary numerical results show that the proposed method is promising.
© 2009 Elsevier Ltd. All rights reserved.
1. Introduction
In this paper, we consider the problem of finding a solution of the nonlinear system of equations
F(x) = 0, (1.1)
where F : R
n
→ R
n
is continuous and monotone. By monotonicity, we mean
hF(x) − F (y), x − yi ≥ 0, ∀x, y ∈ R
n
.
Nonlinear monotone equations have many practical background such as the first order necessary condition of the
unconstrained convex optimization problem and the subproblems in the generalized proximal algorithms with Bregman
distances [1]. Some monotone variational inequality problems can also be converted into systems of nonlinear monotone
equations by means of fixed point maps or normal maps if the underlying function satisfies some coercive conditions [2].
Different methods have been developed for nonlinear systems of equations. Newton’s method and quasi-Newton
methods [3–9] are particularly welcome because of their local superlinear convergence property. However, they are typically
unattractive for large-scale nonlinear systems of equations because they need to solve a linear system using the Jacobian
matrix or an approximation of it.
The spectral gradient method [10] is quite easy to be implemented and very efficient for large-scale unconstrained
optimization problems. Recently, Cruz and Raydan [11] extended the spectral gradient method to solve large-scale nonlinear
systems of equations. They introduced a spectral algorithm (SANE) in [11]. Global convergence is guaranteed by means
of a variation of the nonmonotone strategy of Grippo, Lampariello and Lucidi [12]. La Cruz, Martínez and Raydan [13]
proposed a fully derivative-free SANE algorithm (DF-SANE). Numerical experiments show that DF-SANE works well for
a class of nonlinear systems of equations. Zhang and Zhou [14] combined the spectral gradient method [10] with the
projection method [15] to solve nonlinear monotone equations. The method in [14] is shown to be globally convergent
if the nonlinear equations are Lipschitz continuous. We refer to recent review papers [16,17] for a summary of nonlinear
systems of equations.
I
The work was supported by the NSF project of China granted 10771057.
E-mail address: chengwanyou421@yahoo.com.cn.
0895-7177/$ – see front matter © 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.mcm.2009.04.007