Abstract and Applied Analysis

Volume 2014 (2014), Article ID 131482, 10 pages

http://dx.doi.org/10.1155/2014/131482

## Research on Adjoint Kernelled Quasidifferential

^{1}School of Mathematical Sciences, Dalian University of Technology, Dalian 116024, China^{2}Department of Hydrography & Cartography, PLA Dalian Naval Academy, Dalian 116018, China

Received 10 October 2013; Accepted 19 January 2014; Published 5 March 2014

Academic Editor: Hichem Ben-El-Mechaiekh

Copyright © 2014 Si-Da Lin et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

#### Abstract

The quasidifferential of a quasidifferentiable function in the sense of Demyanov and Rubinov is not uniquely defined. Xia proposed the notion of the kernelled quasidifferential, which is expected to be a representative for the equivalence class of quasidifferentials. Although the kernelled quasidifferential is known to have good algebraic properties and geometric structure, it is still not very convenient for calculating the kernelled quasidifferentials of and , where and are kernelled quasidifferentiable functions. In this paper, the notion of adjoint kernelled quasidifferential, which is well-defined for and , is employed as a representative of the equivalence class of quasidifferentials. Some algebraic properties of the adjoint kernelled quasidifferential are given and the existence of the adjoint kernelled quasidifferential is explored by means of the minimal quasidifferential and the Demyanov difference of convex sets. Under some condition, a formula of the adjoint kernelled quasidifferential is presented.

#### 1. Introduction

Quasidifferential calculus, developed by Demyanov and Rubinov, plays an important role in nonsmooth analysis and optimization. The class of quasidifferentiable functions is fairly broad. It contains not only convex, concave, and differentiable functions but also convex-concave, D.C. (i.e., difference of two convex), maximum, and other functions. In addition, it even includes some functions which are not locally Lipschitz continuous. Quasidifferentiability can be employed to study a wide range of theoretical and practical issues in many fields, such as in mechanics, engineering, and economics nonsmooth analysis and fuzzy control theory (see, e.g., [1–13]).

A function defined on an open set is called quasidifferentiable (q.d.) at a point , in the sense of Demyanov and Rubinov [5], if it is directionally differentiable at and there exist two nonempty convex compact sets and such that the directional derivative can be represented in the form as where denotes the usual inner product in . The pair of sets is called a quasidifferential of at and and are called a subdifferential and a superdifferential, respectively.

It is well known that the quasidifferential is not uniquely defined. Let be the set of all nonempty convex compact sets in . Denote and , where and . Suppose that is a quasidifferential of ; then, for any , the pair of sets is still a quasidifferential of . And the set of quasidifferentials of at is so large that the whole space could be covered by the union of subdifferentials or superdifferentials; that is, The quasidifferential uniqueness is an essential problem in quasidifferential calculus, so it is necessary to find a way by which a quasidifferential, particularly a small quasidifferential in some sense, as a representative of the equivalence class of quasidifferentials, can be determined automatically. The problem was for the first time considered in a discussion at IIASA, by Demyanov and Xia in 1984 [4]. There were many reports and publications mentioning or dealing with this subject from different points of view (see, for instance, [9–26], etc.).

Pallaschke et al. [18] introduced the notion of the minimal quasidifferential and proved the existence of equivalent minimal quasidifferential. is called minimal, provided that satisfying and implies and . Nevertheless, the minimal quasidifferential is not uniquely defined either. Indeed, any translation of a minimal quasidifferential is still a minimal quasidifferential; in other words, if is a minimal quasidifferential, then, for any singleton , the pair of sets is still a minimal quasidifferential. For one-dimensional space, equivalent minimal pairs are uniquely determined up to translations, according to [8]. Grzybowski [15] and Scholtes [22] proved independently the fact that equivalent minimal quasidifferentials, in the two-dimensional case, are uniquely determined up to a translation. For the -dimensional case (), Grzybowski [15] gave the first example of two equivalent minimal pairs in which are not related by translations, and, as in [19], Pallaschke and Unbański indicated that a continuum of equivalent pairs are not related by translation for different indices. Some sufficient conditions and both sufficient and necessary conditions for the minimality of pairs of compact convex sets were given and some reduction techniques for the reduction of pairs of compact convex sets via cutting hyperplanes or excision of compact convex subsets were proposed according to Pallaschke and Urbański [20, 21].

For the same purpose, Xia [24, 25] introduced the notion of the kernelled quasidifferential. It was proved that are nonempty, according to Deng and Gao [14]. and (defined by (3)) are called sub- and super-kernel, respectively, and is called a quasi-kernel of . The quasi-kernel is said to be a kernelled quasidifferential of at if and only if the quasi-kernel is a quasidifferential, denoted by . If has a kernelled quasidifferential at , then is said to be a kernelled quasidifferentiable function at . For the case of one-dimensional space, the existence of the kernelled quasidifferential was given by Gao [16]. In the two dimensional case, based on the translation of minimal quasidifferentials, it was proved that the kernelled quasidifferential exists for any q.d. function (see [17]). In the -dimensional case (), whether the pair of sets given in (3) is a quasidifferential of at is still an open problem, some progress has been made in the last years. Zhang et al. [26] gave a sufficient condition for a quasi-kernel being a kernelled quasidifferential. In [11], Gao presented a condition in terms of Demyanov difference, called g-condition, in which the kernelled quasidifferential exists. The corresponding subclasses and augmented class of g-q.d. functions on were defined and some more properties on this class were presented according to Song and Xia [23].

Although the kernelled quasidifferential is known to have good algebraic properties and geometric structure (see [25]), it is still not very convenient for calculating the kernelled quasidifferentials of and , where and are kernelled quasidifferentiable functions. Hence, in this paper, the notion of adjoint kernelled quasidifferential, which is well-defined for and , is employed as a representative of the equivalence class of quasidifferentials. Some algebraic properties of the adjoint kernelled quasidifferential are given and the existence of the adjoint kernelled quasidifferential is explored by means of the minimal quasidifferential and the Demyanov difference of convex sets. The rest of the paper is organized as follows. In Section 2, some preliminary definitions and results used in the paper are provided. In Section 3, definitions of adjoint kernelled quasidifferential will be introduced and some operations of adjoint kernelled quasidifferentiable functions are given. In Section 4, we prove that the adjoint kernelled quasidifferential exists in one- and two-dimensional cases and two sufficient conditions for the existence of the adjoint kernelled quasidifferential in are given. In Section 5, under some condition, a formula of the adjoint kernelled quasidifferential is presented.

#### 2. Preliminaries

The support function of a set is defined by It is well known (see, e.g., [6]) that the mapping called the Minkowski duality is one-to-one correspondence between and the set of all finite sublinear functions is defined on .

Proposition 1. *Let ; then
*

*It is true that is convex with
particularly, , where denotes the subdifferential in the sense of convex analysis [27].*

*For any and , we denote the max-face of with respect to by the formula
Obviously, the max-face coincides with the subdifferential . Denote by the normal cone to at ; that is,
*

*Proposition 2. Let , for ; it holds
*

*Proposition 3. Let and . If , then
*

*Let the function defined on be locally Lipschitz continuous and let denote the set where exists. The Clarke subdifferential of at is defined as follows:
where “” denotes the convex hull. In the convex case, the Clarke subdifferential coincides with the subdifferential in the sense of convex analysis [28].*

*A set is called of full measure (with respect to ), if is a set of measure zero. Let and be the set of all points such that exists. The set is of full measure in . Let and be a subset of of full measure; then the set
is called Demyanov difference of and , where “cl” refers to the closure. This construction was applied implicitly by Demyanov for the study of connections between the Clarke subdifferential and the quasidifferential [3]. In general, the Demyanov difference is smaller than the Minkowski difference. It is true that
According to [6], the Demyanov difference can be rewritten as
*

*Define the algebraic operations of addition and multiplication by a real number in and the equivalence relation as follows:
where , , and . It is easy to check that .*

*Proposition 4. If , then
*

*The main formulas of quasidifferential calculus will be stated as Proposition 5. Algebraic operations over quasidifferentials are performed as over elements of the space of compact sets (or what is the same, as over pairs of sets).*

*Proposition 5. Let denote the set of all functions defined on an open set and quasidifferentiable at a point . Then, the following hold. (1)If , , are real numbers, then , and Note that, in particular, .(2)Let . Then, and(3)If , , then is quasidifferentiable at and(4)Let and Then, , , and where Here, , .*

*3. Adjoint Kernelled Quasidifferential*

*3. Adjoint Kernelled Quasidifferential**The kernelled quasidifferential is known to have good algebraic properties (see [25]), but it is still not very convenient for calculating the kernelled quasidifferentials of and , where and are kernelled quasidifferentiable functions. So it is natural and necessary to explore the pair of sets , where is defined as in (3) and
Obviously, is nonempty and symmetric. Since having the similar structure to the quasi-kernel of , is called an adjoint quasi-kernel of , where and are called adjoint sub-kernel and adjoint super-kernel, respectively. Of course and are compact convex. This motivates the introduction of the following notions.*

*Definition 6. *Let . The adjoint quasi-kernel is said to be an adjoint kernelled quasidifferential of at if and only if
If has an adjoint kernelled quasidifferential at , then is said to be an adjoint kernelled quasidifferentiable function at . The adjoint kernel is a quasidifferential, denoted by .

*From the definition of quasidifferential and Proposition 5, the following proposition can be obtained immediately, which is especially useful in the study of the operation rules of adjoint kernelled quasidifferential.*

*Proposition 7.
(1) If , , , then
*

Note that, in particular, .

(2) Let . Then, (3) If , , then

*If the adjoint kernelled quasidifferential exists, some operation rules of adjoint kernelled quasidifferential are presented as follows.*

*Theorem 8. Let denote the set of all functions in and having adjoint kernelled quasidifferential at . Then, the following hold. (1)If , then and
(2)If , , then and
(3)If , then and
(4)If , , then and
*

*Proof. *We will prove only Properties (1) and (2). Properties (3) and (4) can be proved in an analogous manner.(1)Since , then
From Propositions 5 and 7 and (32), it follows that
By the similar way, we can prove that
Since , hence, together with (33) and (34), one has that .(2)Since , then, together with Propositions 5 and 7, one has that
Similarly, we can prove that
Combining (35) with (36) leads to
Hence, .

*By we denote the set of all functions in and having kernelled quasidifferential at . Obviously, one has that . The adjoint kernelled quasidifferential is convenient for calculating and can calculate the adjoint kernelled quasidifferential of with kernelled quasidifferential, where , .*

*Theorem 9. If , then and
If , then and
*

*Proof. *Since , then , where
By Propositions 5 and 7 and (41), we obtain
From Propositions 5 and 7 and (40), it follows that
Obviously, . This fact, together with (42) and (43), implies that
Then, and . Similarly, it can be proved that if , then and . The proof is completed.

*Theorem 10. Let and
Then, and , where
Here, .*

*Proof. *Since and , , then, according to Propositions 5 and 7, we have
Since, for , , where denotes a finite index set, one has that
where , . Hence, together with Proposition 5, it follows that
By Propositions 5 and 7 and (49), we obtain

Based on Propositions 5 and 7 and (47) and (50), one has that
hence . The demonstration is completed.

*4. Existence of the Adjoint Kernelled Quasidifferential*

*4. Existence of the Adjoint Kernelled Quasidifferential*

*In this section, the existence of the adjoint kernelled quasidifferential of a quasidifferentiable function is established. In one- and two-dimensional cases, we prove that the adjoint kernelled quasidifferential exists and give its expression by using of a minimal quasidifferential. We also develop the existence of the adjoint kernelled quasidifferential for a quasidifferentiable function on under some conditions.*

*Theorem 11. Suppose that , , and is a minimal quasidifferential of at . Then, the relations below hold
Furthermore, ; that is, .*

*Proof. *Let . From the existence of the minimal quasidifferentials, see [18], it follows that there exists a minimal quasidifferential of at , denoted by , such that , . Consequently,Note that both and are the minimal quasidifferentials of at . According to the translation property of the equivalent minimal quasidifferentials in the one- and two-dimensional case, see [15, 18], there exists , , such that the minimal quasidifferential can be expressed as
This leads toIt follows from (53a), (53b), (55a), and (55b) thatTaking the intersection on the right hands of (56a) and of (56b) for all quasidifferentials of at , we have thatOn the other hand, implies thatThe relations (57a), (57b), (58a), and (58b) lead to thatNote that and , . Hence,
Equations (59a), (59b), and (60) show that . The proof is completed.

*The conclusion of Theorem 11 strongly depends upon the translation of minimal quasidifferentials. Unfortunately, the minimal quasidifferential is not uniquely determined up to a translation in if [15]. But. by the tool of Demyanov difference of compact convex sets, we get the following interesting result about minimal quasidifferential.*

*Proposition 12. Suppose that and there exists a quasidifferential such that
Then is a minimal quasidifferential of at .*

*Proof. *Let and
Obviously, one has
By Proposition 4 and (61), we obtain
From (13) and (64), it follows that
Combining (63) with (65) leads to
According to (62) and (66), we conclude that
Then, by the definition of the minimal quasidifferential, is a minimal quasidifferential of at .

*Inspired by Proposition 12, we present the following theorem, which gives a sufficient condition for the existence of the adjoint kernelled quasidifferential in .*

*Theorem 13. Suppose that and there exists a quasidifferential such that
Then, one hasFurthermore, ; that is, .*

*Proof. *Let . From Proposition 4 and (68), it follows that
By the definition of the quasidifferential, it is easy to check that implies . Therefore, we have and . These give
By (70) and Proposition 1, we obtain
Evidently, (72) is equivalent to the following:
Combining (71) with (73) leads to
Based on (74) and Proposition 1, one has that
Notice that both (70) and (75) hold for any . Taking the intersection on the right-hand sides of (70) and of (75), respectively, for all quasidifferentials of at , it is obtained thatOn the other hand, impliesCombining (76a) with (77b) yields (69a). Likewise, (76b) and (77b) yield (69b). Notice that the relation has been claimed. We thus complete the proof of the theorem.

*A decomposition structure of is defined by
where and are defined by
respectively. Generally, and are positively homogeneous, but not sublinear. It is easy to be seen that
It is easy to be seen that, for any , there exists at least one sequence convergent to , where . According to Proposition 2, if and such that there exist sequences and , then and .*

*The above lines enable us to give the following theorem which provides a sufficient condition for to be an adjoint kernelled quasidifferential.*

*Let be a shape of that is defined by a similar way according to [18], such that
*

*Theorem 14. Let and suppose that and are continuous with respect to direction, and, furthermore, there exists a shape of such that, for any and , one has that
where denotes the closed convex conical hull. If, for any , , and , there exist sequences
such that is one of clusters of ; then , that is, .*

*Proof. *Let be an arbitrary nonzero vector. There exist and such that . According to (82), there exists a sequence
, convergent to . For each , there are two index sets and , with finite indices such that
It follows from (83)–(85) and (87) that, for each , there exist , , , and such that
Since each is a convex combination of , , or of , , one has that there are and such that
satisfying
from (83) and (84), where . Since , , it follows, from the sufficient condition for given before the theorem, that
Thus, it follows from (91) that
Without loss of generality, assume . Taking the limit to (92), one has that
According to the continuity of , (93) becomes
Similarly, it can be proved that
According to (94) and (95), we conclude
Then, by the definition of the quasidifferential, one has , that is, . The demonstration is completed.

*5. Formula of Representative for Quasidifferentials*

*5. Formula of Representative for Quasidifferentials*

*Theorem 13 only gives the existence of the adjoint kernelled quasidifferential but does not show us how to calculate it. For the practical purpose, we expect to find a way to calculate a representative of the equivalent class of quasidifferentials for a given quasidifferential. The present section is devoted to this topic.*

*Lemma 15. Let , . Then, has the following form:
*

*Proof. *Evidently,
This leads to
Taking the Clarke subdifferential at , (99) becomes
Based on the definition of the Demyanov difference, (100) yields ; that is, (97) holds.

*Theorem 16. Let and . If there exists such that , then
*

*Proof. *Setting and in Lemma 15, we have
This completes the proof of the theorem.

*Theorem 17. Let . If there exists satisfying , then, for any , the pair of sets
is the adjoint kernelled quasidifferential of .*

*Proof. *By Theorem 13 and Proposition 4,
is the kernelled quasidifferential. According to Theorem 16, leads to
This means that (103) is the kernelled quasidifferential. The proof is concluded.

*Noticing that the Demyanov difference and the Minkowski difference of polyhedra are polyhedra, we have the following corollary.*

*Corollary 18. Suppose that there exist satisfying and a pair of polyhedra . Then, the kernelled quasidifferential is a pair of polyhedra.*

*Based on above two theorems, given a quasidifferential, the adjoint kernelled quasidifferential can be formulated under some conditions, for instance, the condition in Theorem 13. In particular, if a polyhedral quasidifferential is given, the adjoint kernelled quasidifferential can be calculated because the Demyanov difference of polyhedra can be calculated (for instance, see [9]).*

*Conflict of Interests*

*Conflict of Interests*

*The authors declare that there is no conflict of interests regarding the publication of this paper.*

*Acknowledgment*

*Acknowledgment*

*This paper is supported by the National Natural Science Foundation of China under Grant no. 11171049.*

*References*

*References*

- R. Baier, E. Farkhi, and V. Roshchina, “The directed and Rubinov subdifferentials of quasidifferentiable functions—part I: definition and examples,”
*Nonlinear Analysis: Theory, Methods & Applications*, vol. 75, no. 3, pp. 1074–1088, 2012. View at Publisher · View at Google Scholar · View at MathSciNet - R. Baier, E. Farkhi, and V. Roshchina, “The directed and Rubinov subdifferentials of quasidifferentiable functions—part II: calculus,”
*Nonlinear Analysis: Theory, Methods & Applications*, vol. 75, no. 3, pp. 1058–1073, 2012. View at Publisher · View at Google Scholar · View at MathSciNet - V. F. Demyanov, “On a relation between the Clarke subdifferential and quasidifferential,”
*Vestnik Leningrad University*, vol. 13, pp. 183–189, 1981. View at Google Scholar - V. F. Demyanov and Z. Q. Xia,
*Minimal Quasidifferentials*, International Institute for Applied Systems Analysis (IIASA), Laxenburg, Austria, 1984. - V. F. Demyanov and A. M. Rubinov,
*Quasidifferential Calculus*, Optimization Software, New York, NY, USA, 1986. View at MathSciNet - V. F. Demyanov and A. M. Rubinov,
*Constructive Nonsmooth Analysis*, vol. 7 of*Approximation & Optimization*, Peter Lang, Frankfurt, Germany, 1995. View at MathSciNet - V. F. Dem'yanov, G. E. Stavroulakis, L. N. Polyakova, and P. D. Panagiotopoulos,
*Quasidifferentiability and Nonsmooth Modelling in Mechanics, Engineering and Economics*, vol. 10 of*Nonconvex Optimization and Its Applications*, Kluwer Academic, Dordrecht, The Netherlands, 1996. View at MathSciNet - V. F. Demyanov and A. M. Rubinov,
*Quasidifferentiability and Related Topics*, Kluwer Academic, Dodrecht, The Netherlands, 2000. - Y. Gao, “Demyanov difference of two sets and optimality conditions of Lagrange multiplier type for constrained quasidifferentiable optimization,”
*Journal of Optimization Theory and Applications*, vol. 104, no. 2, pp. 377–394, 2000. View at Publisher · View at Google Scholar · View at MathSciNet - Y. Gao, “Representation of the Clarke generalized Jacobian via the quasidifferential,”
*Journal of Optimization Theory and Applications*, vol. 123, no. 3, pp. 519–532, 2004. View at Publisher · View at Google Scholar · View at Zentralblatt MATH · View at MathSciNet - Y. Gao, “Representative of quasidifferentials and its formula for a quasidifferentiable function,”
*Set-Valued Analysis*, vol. 13, no. 4, pp. 323–336, 2005. View at Publisher · View at Google Scholar · View at Zentralblatt MATH · View at MathSciNet - Y. Gao, “Calculating the proximal subdifferential via the quasidifferential,”
*Applied Mathematics Letters*, vol. 21, no. 11, pp. 1172–1176, 2008. View at Publisher · View at Google Scholar · View at Zentralblatt MATH · View at MathSciNet - B. Luderer and D. Wagner, “Algorithms of quasidifferentiable optimization for the separation of point sets,” in
*Optimization and Optimal Control: Theory and Applications*, vol. 39 of*Springer Optimization and Its Applications*, pp. 157–167, Springer, New York, NY, USA, 2010. View at Publisher · View at Google Scholar · View at Zentralblatt MATH · View at MathSciNet - M. R. Deng and Y. Gao, “A property on quasidifferentials,”
*Chinese Journal of Operations Research*, vol. 10, no. 1, pp. 65–67, 1991 (Chinese). View at Google Scholar - J. Grzybowski, “Minimal pairs of convex compact sets,”
*Archiv der Mathematik*, vol. 63, no. 2, pp. 173–181, 1994. View at Publisher · View at Google Scholar · View at Zentralblatt MATH · View at MathSciNet - Y. Gao, “The star-kernel for a quasidifferentiable function in one dimensional space,”
*Journal of Mathematical Research and Exposition*, vol. 8, no. 1, p. 152, 1988. View at Google Scholar - Y. Gao, Z. Q. Xia, and L. W. Zhang, “Kernelled quasidifferential for a quasidifferentiable function in two-dimensional space,”
*Journal of Convex Analysis*, vol. 8, no. 2, pp. 401–408, 2001. View at Google Scholar · View at Zentralblatt MATH · View at MathSciNet - D. Pallaschke, S. Scholtes, and R. Urbański, “On minimal pairs of convex compact sets,”
*Bulletin of the Polish Academy of Sciences. Mathematics*, vol. 39, no. 1-2, pp. 105–109, 1991. View at Google Scholar · View at Zentralblatt MATH · View at MathSciNet - D. Pallaschke and R. Urbański, “A continuum of minimal pairs of compact convex sets which are not connected by translations,”
*Journal of Convex Analysis*, vol. 3, no. 1, pp. 83–95, 1996. View at Google Scholar · View at Zentralblatt MATH · View at MathSciNet - D. Pallaschke and R. Urbański, “Some criteria for the minimality of pairs of compact convex sets,”
*Zeitschrift für Operations Research*, vol. 37, no. 2, pp. 129–150, 1993. View at Publisher · View at Google Scholar · View at Zentralblatt MATH · View at MathSciNet - D. Pallaschke and R. Urbański, “Reduction of quasidifferentials and minimal representations,”
*Mathematical Programming*, vol. 66, no. 2, pp. 161–180, 1994. View at Publisher · View at Google Scholar · View at Zentralblatt MATH · View at MathSciNet - S. Scholtes, “Minimal pairs of convex bodies in two dimensions,”
*Mathematika*, vol. 39, no. 2, pp. 267–273, 1992. View at Publisher · View at Google Scholar · View at Zentralblatt MATH · View at MathSciNet - C. L. Song and Z. Q. Xia, “A note on the kernelled quasidifferential in the n-dimensional space,”
*Fuzzy Engineering and Operations Research*, vol. 147, pp. 541–547, 2012. View at Google Scholar - Z. Q. Xia,
*The *-Kernel for Quasidifferntiable Functions*, WP-87-89, IIASA, Laxenburg, Austria, 1987. - Z. Q. Xia, “On quasidifferential kernels,”
*Demonstratio Mathematica*, vol. 26, no. 1, pp. 159–182, 1993. View at Google Scholar · View at Zentralblatt MATH · View at MathSciNet - L. W. Zhang, Z. Q. Xia, Y. Gao, and M. Z. Wang, “Star-kernels and star-differentials in quasidifferential analysis,”
*Journal of Convex Analysis*, vol. 9, no. 1, pp. 139–158, 2002. View at Google Scholar · View at Zentralblatt MATH · View at MathSciNet - R. T. Rockafellar,
*Convex Analysis*, Princeton Mathematical Series, Princeton University Press, Princeton, NJ, USA, 1970. View at MathSciNet - F. H. Clarke,
*Optimization and Nonsmooth Analysis*, Canadian Mathematical Society Series of Monographs and Advanced Texts, Wiley-Interscience, New York, NY, USA, 1983. View at MathSciNet

*
*