- Research
- Open Access

# A smoothing and regularization predictor-corrector method for nonlinear inequalities

- Haitao Che
^{1}Email author

**2012**:214

https://doi.org/10.1186/1029-242X-2012-214

© Che; licensee Springer 2012

**Received:**14 July 2011**Accepted:**19 September 2012**Published:**2 October 2012

## Abstract

For a system of nonlinear inequalities, we approximate it by a family of parameterized smooth equations via a new smoothing function. We present a new smoothing and regularization predictor-corrector algorithm. The global and local superlinear convergence of the algorithm is established. In addition, the smoothing parameter *μ* and the regularization parameter *ε* in our algorithm are viewed as different independent variables. Preliminary numerical results show the efficiency of the algorithm.

**MSC:**90C33, 90C30, 15A06.

## Keywords

- nonlinear inequalities
- predictor-corrector algorithm
- ${P}_{0}$-function
- smoothing function

## 1 Introduction

where $f(x)={({f}_{1}(x),{f}_{2}(x),\dots ,{f}_{n}(x))}^{\mathrm{\top}}$ and ${f}_{i}:{R}^{n}\to R$ is a continuously differentiable function for $i=1,2,\dots ,n$. This problem finds applications in data analysis, set separation problems, computer-aided design problems and image reconstructions [1–3]. Among various solution methods for the inequality problems [4–10], the smoothing-type methods receive much attention [8–10] which first transform the problem as a system of nonsmooth equations and approximate it by a smooth equation and then solve it by the smoothing Newton methods. Since the derivative of the underlying mapping may be seriously ill-conditioned, which may prevent the smoothing methods from converging to a solution of the problem, a perturbed regularization technique is introduced to overcome this drawback [9, 11, 12]. In 2003, Huang *et al.* proposed a predictor-corrector smoothing Newton method for nonlinear complementarity problem with a ${P}_{0}$ function based on the perturbed minimum function [13]. The method was shown to be locally superlinear convergent under the assumptions that all $V\in \partial H({z}^{\ast})$ are nonsingular and ${f}^{\mathrm{\prime}}(x)$ is locally Lipschitz continuous around ${x}^{\ast}$.

In this paper, motivated by the smoothed penalty function for constrained optimization [14], we construct a new smoothing function for nonlinear inequalities, and thus we can approximate the nonsmooth system of transformed equations by a system of smooth equations. We develop a regularization smoothing predictor-corrector method for solving the problem by modifying and extending the method in [13]. Besides choosing an arbitrarily starting point, the presented algorithm is simpler than the predictor-corrector noninterior continuation methods developed by Burke and Xu [15].

The rest of this paper is organized as follows. In Section 2, we review some preliminaries to be used in the subsequent analysis and introduce a new smoothing function and its properties. In Section 3, we present a smoothing and regularization predictor-corrector method for solving the nonlinear inequalities and establish the global and local convergence of the proposed algorithm. Preliminary numerical experiments are reported to show the efficiency of the algorithm in Section 5.

To end this section, we introduce some notations used in this paper. The set of $m\times k$ matrices with real entries is denoted by ${R}^{m\times k}$, ${R}_{+}^{n}$ (${R}_{++}^{n}$) denotes the nonnegative (positive) orthant in ${R}^{n}$. The superscript ^{⊤} denotes the transpose of a matrix or a vector. Define $N=\{1,2,\dots ,n\}$, and for any vector $a\in {R}^{n}$, we let ${D}_{a}$ denote the diagonal matrix whose *i*-th diagonal element is ${a}_{i}$. $\parallel u\parallel $ denotes the 2-norm of a vector $u\in {R}^{n}$. For a continuously differentiable function $f:{R}^{n}\to {R}^{m}$, we denote the Jacobian of *f* at $x\in {R}^{n}$ by ${f}^{\prime}(x)$.

## 2 Smooth reformulation of nonlinear inequalities

In this section, we first review some definitions and basic results, and then introduce a new smoothing function and show its properties.

**Definition 2.1** A matrix $M\in {R}^{n\times n}$ is said to be a ${P}_{0}$-matrix if every principle minor of *M* is nonnegative.

**Definition 2.2**A function $F:{R}^{n}\to {R}^{n}$ is said to be a ${P}_{0}$-function if for all $x,y\in {R}^{n}$ with $x\ne y$, there exists an index ${i}_{0}\in N$ such that

For a ${P}_{0}$-matrix, the following conclusion holds [16].

**Lemma 2.1**

*If*$M\in {R}^{n\times n}$

*is a*${P}_{0}$-

*matrix*,

*then every matrix of the form*

*is nonsingular for all positive definite diagonal matrices* ${D}_{a},{D}_{b}\in {R}^{n\times n}$.

**Definition 2.3**Suppose that $G:{R}^{n}\to {R}^{m}$ is a locally Lipschitz function.

*G*is said to be semi-smooth at

*x*if

*G*is directionally differentiable at

*x*and

exists for any $h\in {R}^{n}$, where $\partial G(x)$ denotes the generalized derivative in [17].

The concept of semi-smoothness was originally introduced by Mifflin for functions [18]. Qi and Sun extended the definition of a semi-smooth function to vector-valued functions [19]. Convex functions, smooth functions, piecewise linear functions, convex and concave functions, and sub-smooth functions are examples of semi-smooth functions. A function is semi-smooth at *x* if and only if all its component functions are. The composition of semi-smooth functions is still a semi-smooth function.

**Lemma 2.2** [19]

*Suppose that*$\phi :{R}^{n}\to {R}^{m}$

*is a locally Lipschitz function semi*-

*smooth at x*.

*Then*

- (a)
*for any*$V\in \partial \phi (x+th)$, $h\to 0$,$Vh-{\phi}^{\mathrm{\prime}}(x;h)=o(\parallel h\parallel ),$ - (b)
*for any*$h\to 0$,$\phi (x+h)-\phi (x)-{\phi}^{\mathrm{\prime}}(x;h)=o(\parallel h\parallel ).$

where a smoothing parameter $\mu >0$ and $coshx=\frac{{e}^{x}+{e}^{-x}}{2}$.

This new smoothing function has the following properties.

**Lemma 2.3**

*For any*$(\mu ,a)\in {R}_{++}\times {R}^{n}$,

*it holds that*

- (1)
$\varphi (\cdot ,\cdot )$

*is continuously differentiable at any*$(\mu ,a)\in {R}_{++}\times {R}^{n}$. - (2)
*Let*$\varphi (0,a)={lim}_{\mu \to 0}\varphi (\mu ,a)$,*then*$\varphi (0,a)={a}_{+}$. - (3)
$\frac{\partial \varphi (\mu ,a)}{\partial a}\ge 0$

*at any*$(\mu ,a)\in {R}_{++}\times {R}^{n}$.

*Proof* (1) is straightforward, so we only prove (2) and (3).

Then $\varphi (0,a)={a}_{+}$.

then $\frac{\partial \varphi (\mu ,a)}{\partial a}\ge 0$ at any $(\mu ,a)\in {R}_{++}\times {R}^{n}$. We complete the proof. □

**Theorem 2.1**

*Let*$H(\mu ,\epsilon ,x)$

*be defined as*(2.3).

*Then*

- (a)$H(\mu ,\epsilon ,x)$
*is continuously differentiable at any*$z=(\mu ,\epsilon ,x)\in {R}_{++}\times {R}_{++}\times {R}^{n}$*with its Jacobian*${H}^{\mathrm{\prime}}(z)=\left(\begin{array}{ccc}1& 0& 0\\ 0& 1& 0\\ {\mathrm{\Phi}}_{\mu}^{\mathrm{\prime}}(z)& x& {\mathrm{\Phi}}_{x}^{\mathrm{\prime}}(z)\end{array}\right),$(2.7)

*where*

- (b)
*If**f**is a*${P}_{0}$-*function*,*then*${H}^{\mathrm{\prime}}(z)$*is nonsingular at any*${R}_{++}\times {R}_{++}\times {R}^{n}$.

*Proof* (a) is straightforward, so we only prove (b). For (b), we only need to show ${\mathrm{\Phi}}_{x}^{\mathrm{\prime}}(z)$ is nonsingular. In fact, since *f* is a ${P}_{0}$-function, then ${f}^{\mathrm{\prime}}(x)$ is a ${P}_{0}$-matrix for all $x\in {R}^{n}$ by Theorem 3.3 in [22]. We also note that $diag\{(1+\frac{sinh\frac{{f}_{i}(x)}{\mu}}{1+cosh\frac{{f}_{i}(x)}{\mu}}):i\in N\}$ and *εI* are positive diagonal matrices, we know that ${\mathrm{\Phi}}_{x}^{\mathrm{\prime}}(z)$ is nonsingular by Lemma 2.1, which implies that ${H}^{\mathrm{\prime}}(z)$ is also nonsingular. This completes the proof. □

## 3 Algorithm and convergence

In this section, we first describe our algorithm and then we reveal the global convergence analysis of the algorithm. Now, we are at a position to give the description of our smoothing predictor-corrector algorithm.

**Algorithm 3.1**

Step 0. Take $\delta \in (0,1)$, $\sigma \in (0,1)$. Let ${e}^{0}=({\mu}_{0},{\epsilon}_{0},0)\in {R}_{++}\times {R}_{++}\times {R}^{n}$ and ${x}^{0}\in {R}^{n}$ is an arbitrary point. Choose ${z}^{0}=({\mu}_{0},{\epsilon}_{0},{x}^{0})$ and parameter $\gamma \in (0,1)$ such that $\gamma \parallel H({z}^{0})\parallel \le 1$, $\gamma {\mu}_{0}+\gamma {\epsilon}_{0}<1$. Set $k=0$.

Step 1. If $\parallel H({z}^{k})\parallel =0$, then stop. Otherwise, let ${\beta}_{k}=\beta ({z}^{k})$ where $\beta (z)$ is defined by $\beta (z)=\gamma \parallel H(z)\parallel $.

then set ${\stackrel{\u02c6}{z}}^{k}={z}^{k}+\mathrm{\Delta}{\stackrel{\u02c6}{z}}^{k}$. Otherwise, set ${\stackrel{\u02c6}{z}}^{k}={z}^{k}$.

*l*such that

Set ${\lambda}_{k}={\delta}^{{l}_{k}}$ and ${z}^{k+1}:={\stackrel{\u02c6}{z}}^{k}+{\lambda}_{k}\mathrm{\Delta}{\tilde{z}}^{k}$.

Step 4. Set $k:=k+1$ and return to Step 1.

**Remark 3.1** If $\parallel H({z}^{k})\parallel \ge 1$, then Algorithm 3.1 solves only one linear system of equations at each iteration. Otherwise, it solves two linear systems of equations at each iteration. However, the coefficient matrices of these two systems are identical when (3.2) is not satisfied. There are the same points as the algorithm in [13], the neighborhood of the path does not appear in the algorithm, thus, it does not need a few additional computations which keep the iteration sequence staying in the given neighborhood.

The following lemmas show that Algorithm 3.1 is well defined and generates an infinite sequence with some good features.

**Lemma 3.1** *If* *f* *is a continuously differentiable* ${P}_{0}$-*function*, *then Algorithm * 3.1 *is well defined*. *In addition*, ${\mu}_{k}>0$, ${\epsilon}_{k}>0$ *and* ${z}^{k}=({\mu}_{k},{\epsilon}_{k},{x}^{k})\in \mathrm{\Omega}$ *for any* $k\ge 0$.

*Proof*Since

*f*is a continuously differentiable ${P}_{0}$ function, then it follows from Theorem 2.1 that the matrix ${H}^{\mathrm{\prime}}(z)$ is nonsingular for $u>0$, $\epsilon >0$. Since ${u}_{0}>0$, ${\epsilon}_{0}>0$ by the choice of an initial point, we may assume, without loss of generality, that ${\mu}_{k}>0$, ${\epsilon}_{k}>0$, we show that ${\stackrel{\u02c6}{\mu}}_{k}>0$, ${\stackrel{\u02c6}{\epsilon}}_{k}>0$. If the predictor step is accepted, then by (3.1),

otherwise, we have ${z}^{k}={\stackrel{\u02c6}{z}}^{k}$, which means ${\mu}_{k}={\stackrel{\u02c6}{\mu}}_{k}$, ${\epsilon}_{k}={\stackrel{\u02c6}{\epsilon}}_{k}$. Thus, we obtain ${\stackrel{\u02c6}{\mu}}_{k}>0$, ${\stackrel{\u02c6}{\epsilon}}_{k}>0$. Furthermore, ${H}^{\mathrm{\prime}}({z}^{k})$ and ${H}^{\mathrm{\prime}}({\stackrel{\u02c6}{z}}^{k})$ are nonsingular which means that (3.1) and (3.3) are well defined.

That is, the nonnegative ${l}_{k}$ satisfying (3.4) can be found, which demonstrates that (3.4) is well defined.

Similarly, we can obtain ${\epsilon}_{k+1}-{\beta}_{k+1}{\epsilon}_{0}\ge 0$. Thus, ${z}^{k+1}\in \mathrm{\Omega}$.

Since ${u}_{0}>0$, ${\epsilon}_{0}>0$, we may assume that ${\mu}_{k}>0$, ${\epsilon}_{k}>0$ for any given $k\ge 0$. From ${\stackrel{\u02c6}{\mu}}_{k}>0$, ${\stackrel{\u02c6}{\epsilon}}_{k}>0$, it follows from (3.13) that ${\mu}_{k+1}>0$, ${\epsilon}_{k+1}>0$. Hence, ${\mu}_{k}>0$, ${\epsilon}_{k}>0$ for any $k\ge 0$. □

**Lemma 3.2** *Suppose that the infinite sequence* $\{{z}^{k}=({\mu}_{k},{\epsilon}_{k},{x}^{k})\}$ *is generated by Algorithm * 3.1, *then* $0<{\mu}_{k+1}\le {\mu}_{k},0<{\epsilon}_{k+1}\le {\epsilon}_{k}$ *and the sequence* $\{\parallel H({z}^{k})\parallel \}$ *is monotonically decreasing*.

*k*-th iterate, then (3.17) and (3.18) show the desired result. Otherwise, from (3.5), (3.6), $\parallel H({z}^{k})\parallel <1$ and ${z}^{k}\in \mathrm{\Omega}$, one has

Thus, we obtain that ${\mu}_{k+1}\le {\mu}_{k}$, ${\epsilon}_{k+1}\le {\epsilon}_{k}$ hold for any $k\ge 0$.

*k*-th iterate, then (3.15) implies that

which means the sequence $\{\parallel H({z}^{k})\parallel \}$ is monotonically decreasing. □

**Lemma 3.3**

*Assume that*

*f*

*is a*${P}_{0}$-

*function and*${\mu}_{1}$, ${\mu}_{2}$, ${\epsilon}_{1}$, ${\epsilon}_{2}$

*are given positive numbers satisfying*${\mu}_{1}<{\mu}_{2}$, ${\epsilon}_{1}<{\epsilon}_{2}$.

*Then*,

*H*

*defined by*(2.3)

*has the property*

*for any sequence* $\{({\mu}_{k},{\epsilon}_{k},{x}^{k})\}$ *such that* ${\mu}_{1}\le {\mu}_{k}\le {\mu}_{2}$, ${\epsilon}_{1}\le {\epsilon}_{k}\le {\epsilon}_{2}$ *for any* *k* *and* $\parallel {x}^{k}\parallel \to +\mathrm{\infty}$ *as* $k\to +\mathrm{\infty}$.

*Proof*We outline the proof by contradiction. Suppose that the lemma is not true. Then there exists a sequence $\{{z}^{k}=({\mu}_{k},{\epsilon}_{k},{x}^{k})\}$ such that ${\mu}_{1}\le {\mu}_{k}\le {\mu}_{2}$, ${\epsilon}_{1}\le {\epsilon}_{k}\le {\epsilon}_{2}$, $\psi ({z}^{k})\le \psi ({z}^{0})$ but $\parallel {x}^{k}\parallel \to \mathrm{\infty}$. Since the sequence $\{{x}^{k}\}$ is unbounded, the index set $I=\{i\in N:\{{x}_{i}^{k}\}\text{is unbounded}\}$ is nonempty. Without loss of generality, we can assume that $\{|{x}_{i}^{k}|\}\to +\mathrm{\infty}$ for all $i\in I$. Then the following sequence $\{{\overline{x}}^{k}\}$ is bounded which is defined by

*f*is a ${P}_{0}$-function, by Definition 2.2, we have

where ${i}_{0}$ is one of the indices for which the max is attained, and ${i}_{0}$ is assumed, without loss of generality, to be independent of *k*. Since ${i}_{0}\in I$, one has $\{|{x}_{{i}_{0}}^{k}|\}\to +\mathrm{\infty}$ as $k\to \mathrm{\infty}$. We now break up the proof into two cases.

Case 1. If ${x}_{{i}_{0}}^{k}\to +\mathrm{\infty}$ as $k\to \mathrm{\infty}$. In this case, since ${f}_{{i}_{0}}({\overline{x}}^{k})$ is bounded, we deduce from (3.21) that ${f}_{{i}_{0}}({x}^{k})>{f}_{{i}_{0}}({\overline{x}}^{k})$.

Thus, $\parallel \mathrm{\Phi}({z}^{k})\parallel \to +\mathrm{\infty}$ as $k\to \mathrm{\infty}$.

Thus, $\parallel \mathrm{\Phi}({z}^{k})\parallel \to +\mathrm{\infty}$ as $k\to \mathrm{\infty}$.

Case 2. ${x}_{{i}_{0}}^{k}\to -\mathrm{\infty}$ as $k\to \mathrm{\infty}$. In this case, since ${f}_{{i}_{0}}({\overline{x}}^{k})$ is bounded, we deduce from (3.21) that ${f}_{{i}_{0}}({x}^{k})<{f}_{{i}_{0}}({\overline{x}}^{k})$.

Thus, $\parallel \mathrm{\Phi}({z}^{k})\parallel \to +\mathrm{\infty}$ as $k\to \mathrm{\infty}$.

Thus, $\parallel \mathrm{\Phi}({z}^{k})\parallel \to +\mathrm{\infty}$ as $k\to \mathrm{\infty}$.

In summary, we obtain $\psi ({z}^{k})\to +\mathrm{\infty}$ as $k\to \mathrm{\infty}$, which contradicts $\psi ({z}^{k})\le \psi ({z}^{0})$, and the proof is completed. □

*f*being a ${P}_{0}$-function, Lemma 3.2 and Lemma 3.3 indicate that the level set ${L}_{\mu}({z}^{0})$ defined by

is bounded.

To obtain the global convergence of Algorithm 3.1, we need the following assumption.

**Assumption 3.1** The solution $S:=\{x\in {R}^{n},f(x)\le 0\}$ of (1.1) is nonempty and bounded.

Note that Assumption 3.1 seems to be the weakest condition used in the previous literature to ensure the bound of iteration sequences (see [23]).

**Theorem 3.1**

*Assume that the infinite sequence*$\{{z}^{k}\}$

*is generated by Algorithm*3.1.

*Then*

- (a)
*The sequences*$\{\parallel H({z}^{k})\parallel \}$, $\{{\mu}_{k}\}$*and*$\{{\epsilon}_{k}\}$*converge to zero as*$k\to +\mathrm{\infty}$,*and hence any accumulation point of*$\{{x}^{k}\}$*is a solution of*(1.1). - (b)
*If Assumption*4.1*is satisfied*,*then the sequence*$\{{z}^{k}\}$*is bounded*,*hence there exists at least one accumulation point*${z}^{\ast}=({\mu}_{\ast},{\epsilon}_{\ast},{x}^{\ast})$*with*$H({z}^{\ast})=0$*and*${x}^{\ast}\in S$.

*Proof*By Lemma 3.2, we know that $\{\parallel H({z}^{k})\parallel \}$ converges to ${h}^{\ast}$ as $k\to \mathrm{\infty}$. Suppose that $\{\parallel H({z}^{k})\parallel \}$ does not converge to zero. Then, ${h}^{\ast}>0$ and $\{{z}^{k}\}$ is bounded by Lemma 3.2 and Lemma 3.3. Assume that ${z}^{\ast}=({\mu}_{\ast},{\epsilon}_{\ast},{x}^{\ast})$ is an accumulation point of $\{{z}^{k}\}$. Without loss of generality, we assume that $\{{z}^{k}\}$ converges to ${z}^{\ast}$. Then, by the continuity of

*H*and the definition of $\beta (\cdot )$, we know that $\{{\mu}_{k}\}$, $\{{\epsilon}_{k}\}$ and $\{{\beta}_{k}\}$ converge to ${\mu}_{\ast}$, ${\epsilon}_{\ast}$, ${\beta}_{\ast}$ respectively and that ${h}^{\ast}=\parallel H({z}^{\ast})\parallel >0$. Therefore, by (3.4), we have

*i.e.*,

*i.e.*,

This contradicts the fact that $\sigma \in (0,1)$ and $\gamma ({\mu}_{0}+{\epsilon}_{0})<1$. Hence, we have ${h}^{\ast}=0$, ${\mu}_{\ast}=0,{\epsilon}_{\ast}=0$. Thus, $H({z}^{\ast})=0$, that is, ${x}^{\ast}$is a solution of (1.1).

Therefore, by the famous mountain pass theorem (Theorem 5.4 in [24]) and along the lines of the proof of Theorem 3.1 in [23], we obtain that $\{{x}^{k}\}$ is bounded and hence $\{{z}^{k}\}$ is. Thus, $\{{z}^{k}\}$ has at least one accumulation point ${z}^{\ast}=({\mu}_{\ast},{\epsilon}_{\ast},{x}^{\ast})$. By (a), we have $H({z}^{\ast})=0$ and ${\mu}_{\ast}=0$, ${\epsilon}_{\ast}=0$, ${x}^{\ast}\in S$. □

Next, we show the local superlinear convergence of Algorithm 3.1.

**Theorem 3.2**

*Suppose that*

*f*

*is a continuously differentiable*${P}_{0}$-

*function*,

*Assumption*3.1

*is satisfied and*${z}^{\ast}$

*is an accumulation point of the iteration sequence*$\{{z}^{k}\}$

*generated by Algorithm*3.1.

*If all*$V\in \partial H({z}^{\ast})$

*are nonsingular and*${f}^{\mathrm{\prime}}(x)$

*is locally Lipschitz continuous around*${x}^{\ast}$,

*then the whole sequence*$\{{z}^{k}\}$

*superlinearly converges to*${z}^{\ast}$,

*i*.

*e*.,

*and*

*Proof*First, from Theorem 3.1, we know that ${z}^{\ast}$ is a solution of $H(z)=0$. Then since all $V\in \partial H({z}^{\ast})$ are nonsingular, it follows from [19, 25, 27] that for all ${z}^{k}$ sufficiently close to ${z}^{\ast}$, we have

where $C>0$ is a constant.

holds.

*k*is sufficiently large, then ${z}^{k+1}={z}^{k}+\mathrm{\Delta}{z}^{k}$, so

*k*sufficiently large,

This means that ${\mu}_{k+1}=o({\mu}_{k})$, ${\epsilon}_{k+1}=o({\epsilon}_{k})$ and the desired result follows. □

## 4 Numerical experiments

*H*defined by (2.3) is replaced by

where *c* is a constant. It is easy to see that such a change does not destroy any theoretical results obtained in Section 3.

In our numerical experiments, the parameters used in the algorithm are chosen as follows: $\sigma =0.06$, $\delta =0.3$, ${\mu}_{0}=1$, $\gamma =0.01min\{1,1/\parallel H({z}^{0})\parallel \}$. The algorithm terminates when $\parallel \psi ({z}^{k})\parallel \le {10}^{-3}$. In the tables of test results, *st* denotes the starting point of ${x}^{0}$, *ic* denotes the corrector iteration numbers in Step 3 followed directly from Step 1, *ip* denotes the predictor iteration numbers, $iter$ denotes the iteration numbers of smoothing method (in [8]), $cpu$ denotes the CPU time for solving the underlying problems in seconds, and $sol$ denotes a solution of the test problem. In the following, we reveal a detailed description of the tested problems.

**Numerical results of Example 4.1**

Our proposed algorithm | Smoothing algorithm [8] | ||||||||
---|---|---|---|---|---|---|---|---|---|

st | c | ${\mathit{\epsilon}}_{\mathbf{0}}$ | ic | ip | sol | cpu | iter | sol | cpu |

${(0,0)}^{\mathrm{\top}}$ | 10 | 0.5 | 1 | 2 | ${(0.0424,-1.0101)}^{\mathrm{\top}}$ | 0.14 | 23 | ${(0.0592,-0.9961)}^{\mathrm{\top}}$ | 1.1 |

${(1,-1)}^{\mathrm{\top}}$ | 10 | 0.4 | 1 | 2 | ${(0.5090,-0.8655)}^{\mathrm{\top}}$ | 0.18 | 26 | ${(0.3440,-0.9392)}^{\mathrm{\top}}$ | 1.3 |

${(1,-1)}^{\mathrm{\top}}$ | 10 | 0.8 | 1 | 2 | ${(0.6132,-0.7744)}^{\mathrm{\top}}$ | 0.18 | 28 | ${(0.2568,-0.9667)}^{\mathrm{\top}}$ | 1.4 |

**Numerical results of Example 4.2**

Our proposed algorithm | Smoothing algorithm [8] | ||||||||
---|---|---|---|---|---|---|---|---|---|

st | c | ${\mathit{\epsilon}}_{\mathbf{0}}$ | ic | ip | sol | cpu | iter | sol | cpu |

${(0,0)}^{\mathrm{\top}}$ | 0.5 | 1 | 2 | 1 | ${(0.0442,0.3356)}^{\mathrm{\top}}$ | 0.12 | fail | fail | fail |

${(0,0)}^{\mathrm{\top}}$ | 0.5 | 0.5 | 2 | 1 | ${(-0.0105,0.9541)}^{\mathrm{\top}}$ | 0.15 | 21 | ${(-0.0000,1.2045)}^{\mathrm{\top}}$ | 0.87 |

${(1,1)}^{\mathrm{\top}}$ | 0.5 | 0.1 | 2 | 1 | ${(-0.0019,1.5663)}^{\mathrm{\top}}$ | 0.16 | 21 | ${(0.0004,1.5704)}^{\mathrm{\top}}$ | 0.87 |

${(1,1)}^{\mathrm{\top}}$ | 0.5 | 1 | 2 | 1 | ${(-0.0206,0.8605)}^{\mathrm{\top}}$ | 0.17 | 22 | ${(0.0006,1.5698)}^{\mathrm{\top}}$ | 0.90 |

**Numerical results of Example 4.3**

Our proposed algorithm | Smoothing algorithm [8] | ||||||||
---|---|---|---|---|---|---|---|---|---|

st | c | ${\mathit{\epsilon}}_{0}$ | ic | ip | sol | cpu | iter | sol | cpu |

${(0,1)}^{\mathrm{\top}}$ | 10 | 0.5 | 1 | 2 | ${(0.1237,0.5902)}^{\mathrm{\top}}$ | 0.17 | fail | fail | fail |

${(1,1)}^{\mathrm{\top}}$ | 10 | 1 | 1 | 2 | ${(0.0590,0.5236)}^{\mathrm{\top}}$ | 0.16 | 23 | ${(0.5023,0.5153)}^{\mathrm{\top}}$ | 1.26 |

${(1,1)}^{\mathrm{\top}}$ | 1 | 1 | 1 | 2 | ${(0.4533,0.4973)}^{\mathrm{\top}}$ | 0.20 | 23 | ${(0.5274,0.5080)}^{\mathrm{\top}}$ | 1.22 |

**Example 4.3** [26]

## 5 Conclusion

In this paper, we present a new smoothing and regularization predictor-corrector algorithm to solve the nonlinear inequalities, the global and local convergence are obtained. Furthermore, the smoothing parameter *μ* and the regularization parameter *ε* in our algorithm are viewed as independent variables. Preliminary numerical results show the efficiency of the algorithm.

## Declarations

### Acknowledgements

This research was supported by the Natural Science Foundation of China (Grant Nos. 11171180, 11171193, 11126233, 10901096) and the fund of Natural Science of Shandong Province (Grant Nos. ZR2009AL019, ZR2011AM016). The authors are in debt to the anonymous referees for their numerous insightful comments and constructive suggestions which help improve the presentation of the article. The authors thank Prof. Yiju Wang for his careful reading of the manuscript.

## Authors’ Affiliations

## References

- Zakian V, Nail UA: Design of dynamical and control systems by the method of inequalities.
*Proc. Inst. Electr. Eng.*1973, 120: 1421–1472.View ArticleGoogle Scholar - Dennis JE, Schnabel RB:
*Numerical Methods for Unconstrained Optimization and Nonlinear Equations*. Prentice-Hall, Englewood Cliffs; 1983.Google Scholar - Neumaier A:
*Interval Methods for System of Equations*. Cambridge University Press, Cambridge; 1990.Google Scholar - Daniel JW: Newtons method for nonlinear inequalities.
*Numer. Math.*1973, 21: 381–387.MathSciNetView ArticleGoogle Scholar - Mayne DQ, Polak E, Heunis AJ: Solving nonlinear inequalities in a finite number of iterations.
*J. Optim. Theory Appl.*1981, 33: 207–221.MathSciNetView ArticleGoogle Scholar - Sahba M: On the solution of nonlinear inequalities in a finite number of iterations.
*Numer. Math.*1985, 46: 229–236.MathSciNetView ArticleGoogle Scholar - Yin HX, Huang ZH, Qi L: The convergence of a Levenberg-Marquardt method for nonlinear inequalities.
*Numer. Funct. Anal. Optim.*2008, 29: 687–716.MathSciNetView ArticleGoogle Scholar - Huang ZH, Zhang Y, Wu W: A smoothing-type algorithm for solving system of inequalities.
*J. Comput. Appl. Math.*2008, 220: 355–363.MathSciNetView ArticleGoogle Scholar - Zhu JG, Liu HW, Li XL:A regularized smoothing-type algorithm for solving a system of inequalities with a ${P}_{0}$-function.
*J. Comput. Appl. Math.*2010, 233: 2611–2619.MathSciNetView ArticleGoogle Scholar - He C, Ma CF: A smoothing self-adaptive Levenberg-Marquardt algorithm for solving system of nonlinear inequalities.
*Appl. Math. Comput.*2010, 216: 3056–3063.MathSciNetView ArticleGoogle Scholar - Huang ZH, Qi L, Sun D:Sub-quadratic convergence of a smoothing Newton algorithm for the ${P}_{0}$ and monotone LCP.
*Math. Program.*2004, 99: 423–441.MathSciNetView ArticleGoogle Scholar - Zhao N, Huang ZH:A nonmonotone smoothing Newton algorithm for solving box constrained variational inequalities with a ${P}_{0}$ function.
*J. Ind. Manag. Optim.*2011, 7(2):467–482.MathSciNetView ArticleGoogle Scholar - Huang ZH, Han J, Chen Z:Predictor-corrector smoothing Newton method, based on a new smoothing function, for solving the nonlinear complementarity with a ${P}_{0}$ function.
*J. Optim. Theory Appl.*2003, 117: 39–68.MathSciNetView ArticleGoogle Scholar - Herty M, Klar A, Singh AK, Spellucci P: Smoothed penalty algorithms for optimization of nonlinear models.
*Comput. Optim. Appl.*2007, 37: 157–176.MathSciNetView ArticleGoogle Scholar - Burke J, Xu S: A noninterior predictor-corrector path-following algorithm for the monotone linear complementarity problem.
*Math. Program.*2000, 87: 113–130.MathSciNetGoogle Scholar - Luca TD, Facchinei F, Kanzow C: A semismooth equation approach to the solution of nonlinear complementarity problems.
*Math. Program.*1996, 75: 407–439.Google Scholar - Clarke FH:
*Optimization and Nonsmooth Analysis*. Wiley, New York; 1983.Google Scholar - Mifflin R: Semismooth and semiconvex functions in constrained optimization.
*SIAM J. Control Optim.*1977, 15(6):957–972.MathSciNetView ArticleGoogle Scholar - Qi L, Sun J: A nonsmooth version of Newton’s method.
*Math. Program.*1993, 58(3):353–367.MathSciNetView ArticleGoogle Scholar - Lee YJ, Mangasarin OL: SSVM: A smooth support vector machine for classification.
*Comput. Optim. Appl.*2001, 20: 5–22.MathSciNetView ArticleGoogle Scholar - Che H, Li M: A smoothing and regularization Broyden-like method for nonlinear inequalities.
*J. Appl. Math. Comput.*2012. doi:10.1007/s12190–012–0588–2Google Scholar - Chen B, Harker PT: A non-interior-point continuation method for linear complementarity problems.
*SIAM J. Matrix Anal. Appl.*1993, 14: 1168–1190.MathSciNetView ArticleGoogle Scholar - Huang ZH, Han J, Xu D, Zhang L: The non-linear continuation methods for solving the P0-functions non-linear complementarity problem.
*Sci. China*2001, 44(2):1107–1114.MathSciNetView ArticleGoogle Scholar - Facchinei F, Kanzow C: Beyond monotonicity in regularization methods for nonlinear complementarity problems.
*SIAM J. Control Optim.*1999, 37(2):1150–1161.MathSciNetView ArticleGoogle Scholar - Qi L: Convergence analysis of some algorithms for solving nonsmooth equations.
*Math. Oper. Res.*1993, 18: 227–244.MathSciNetView ArticleGoogle Scholar - Zhang Y, Huang ZH: A nonmonotone smoothing-type algorithm for solving a system of equalities and inequalities.
*J. Comput. Appl. Math.*2010, 233: 2312–2321.MathSciNetView ArticleGoogle Scholar - Qi L, Sun DF, Zhou GL: A new look at smoothing Newton methods for nonlinear complementarity problems and box constrained variational inequalities.
*Math. Program., Ser. A*2000, 87: 1–35.MathSciNetGoogle Scholar

## Copyright

This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.