Maximum efficiency for a family of Newton-like methods with frozen derivatives and some applications
Highlights
► We explore the applications of a generalized k-step iterative Newton’s method with frozen derivative. ► We analyze the order and the convergence of the family. ► We are able to compute the maximum computational efficiency in the family for a given example. ► In most of the cases, the 2-step iterative method seems the most efficient.
Introduction
In this paper, a family of Newton-like methods with frozen derivatives is analyzed. This type of method appears in many applications where the authors heuristically choose a given number of steps with frozen derivatives (see for instance this incomplete list of Refs. [14], [21], [22], [23], [24], [25]). Our goal is to give a rigorous estimate of the maximum efficiency of these schemes for a given problem.
Let be a nonlinear operator and assume that F has at least bounded third-order derivatives with continuity on an open convex set D. Suppose that the equation has a solution at which is nonsingular.
Setting , we develop and its derivatives in a neighborhood of . That is, assuming that exists, we havewhere and . From (1), the derivative of can be written as , and expanding the inverse of the derivative of in terms of e in formal power developments, we get
A direct computation of the local order of convergence for Newton’s method and for its generalization in k steps are presented in Section 2. The optimal number of steps in order to maximize the computational efficiency index is computed in Section 3. In Section 4, we analyze several researchers’ examples that illustrate our theoretical results. Section 5 is devoted to the local and semilocal convergence of these schemes. The last section shows an application related to image denoising, where we approximate the total variation (TV) minimization model using the described schemes. A nonlinear primal–dual method is used to remove some of the singularity caused by the nondifferentiability of the quantity in the definition of the TV norm. A finite difference scheme is then applied and the associated nonlinear system of equations is approximated by the most efficient candidate of our family for this particular problem.
Section snippets
Order of convergence
In this section we derive the local order of convergence. This derivation is clearer and more compact than other approaches in the literature for other similar Newton-type methods in several dimensions.
Given , we consider the following iterative method (known as the ‘simplified Newton method’ [1]);where and in the last step, the last computed term is . For , subtracting the root from both sides of (3) and taking into account (1), (2)
Optimal computational efficiency
The computational efficiency index () and the computational cost () are defined as
In (5), and represent the number of evaluations of the scalar functions of F and respectively. The function represents the number of products needed per iteration. The ratios and between products and evaluations are introduced in order to express the value of only in terms of products.
Without
First numerical results
The first set of numerical computations were performed using the MAPLE computer algebra system with 2048 digits. The classical stopping criterionwas used, where is replaced by
and . Note that (9) is independent of knowledge of the root (see [5]).
According to the definition of the computational cost (5), an estimation of the factors is claimed. To do this, we express the cost of the evaluation of the elementary
Local and semilocal analysis of convergence
In this section we propose two convergence results that we will apply to an image denoising problem in the next section. We can find this type of result for the classical Newton method in [9], [10].
On the application of the iterative methods in image processing
During some phases of the manipulation of an image, some random noise and blurring is usually introduced. This noise and blurring makes the later phases of processing the image difficult and inaccurate. In the past two decades many authors have introduced and analyzed certain tools for the image restoration problem. See for instance the following incomplete list; [11], [12], [13], [15], [16], [17], [18], [19], [20], [26], [27], [28].
In this paper, we focus on approaches which use partial
Conclusion
In this paper we have presented a rigorous and useful procedure to find the best scheme in a family of Newton type methods with frozen derivatives for solving a given system of nonlinear equations. These methods have been used in many applications where the authors heuristically chose a given number of steps with frozen derivatives. We have applied this new strategy of this paper to the approximation of several known problems. We have analyzed the order, the computational efficiency and the
References (31)
- et al.
On some computational orders of convergence
Appl. Math. Lett.
(2010) - et al.
A variant of Newton’s method with accelerated third-order convergence
Appl. Math. Lett.
(2000) - et al.
Variational image restoration by means of wavelets: simultaneous decomposition, deblurring, and denoising
Appl. Comput. Harmon. Anal.
(2005) - et al.
Image decompositions using bounded variation and generalized homogeneous Besov spaces
Appl. Comput. Harmon. Anal.
(2007) - et al.
A modified frozen Newton method to identify a cavity by means of boundary measurements
Math. Comput. Simulat.
(2004) - et al.
Nonlinear total variation based noise removal algorithms
Physica D
(1992) - et al.
Iterative Solution of Nonlinear Equations in Several Variables
(1970) Iterative Methods for the Solution of Equations
(1964)On a modification of the Newton method
Ukrain. Math. Ž
(1967)- et al.
Nondiscrete Induction and Iterative Processes
(1984)
Accelerated iterative methods for finding solutions of a system of nonlinear equations
Appl. Math. Comput.
A unified approach for enlarging the radius of convergence for Newton’s method and applications
Nonlinear Funct. Anal. Appl.
Relaxing the convergence conditions for Newton-like methods
J. Appl. Math. Comput.
Analysis of a New Nonlinear Subdivision Scheme. Applications in Image Processing
Foundations of Computational Mathematics
Cited by (0)
- 1
Research supported by MICINN-FEDER MTM2010-17508 (Spain), and by 08662/PI/08 (Murcia).