∎
[email protected]
Rong Hu
[email protected]
Yaping Fang, Corresponding author
[email protected]
1College of Mathematics, Sichuan University, Chengdu, Sichuan, P.R. China
2College of Applied Mathematics, Chengdu University of Information Technology, Chengdu, Sichuan, P.R. China
Solvability of a Regular Polynomial Vector Optimization Problem without Convexity††thanks: This work was partially supported by the National Natural Science Foundation of China (11471230).
Abstract
In this paper we consider the solvability of a non-convex regular polynomial vector optimization problem on a nonempty closed set. We introduce regularity conditions for the polynomial vector optimization problem and study properties and characterizations of the regularity conditions. Under the regularity conditions, we study nonemptiness and boundedness of the solution sets of the problem. As a consequence, we establish two Frank-Wolfe type theorems for the non-convex polynomial vector optimization problem. Finally, we investigate the solution stability of the non-convex regular polynomial vector optimization problem.
Keywords:
Polynomial vector optimization problem Regularity condition Frank-Wolfe type theorem Non-convexity StabilityMathematics Subject Classification (2020) 90C29; 90C23; 49K40.
1 Introduction
Throughout, denotes the -dimensional Euclidean space with the norm and the inner , and . In this paper we are interested in the following polynomial vector optimization problem on :
where is a vector polynomial such that each component function is a polynomial with its degree , and is a nonempty closed set (not necessarily convex set or semi-algebraic set HHV ; BR ).
Recall that a point is a Pareto efficient solution of if
and is a weak Pareto efficient solution of PVOP if
The Pareto efficient solution set and the weak Pareto efficient solution set of are denoted by and respectively. Clearly, . When , collapses to the polynomial scalar optimization problem:
whose solution set is denoted by .
For the polynomial scalar optimization problem, regularity condition has been used in AQ to investigate the nonemptiness of the solution set and the continuity of the solution mapping for a quadratic programming problem. Hieu HV1 established a Frank-Wolfe type theorem for a polynomial scalar optimization problem on a nonempty closed set by proving the nonemptiness and boundedness of its solution set when the objective function is bounded from below on the constraint set and the regularity condition holds. Hieu et al. HV2 proved that the solution set of an optimization problem corresponding to a polynomial complementarity problem is nonempty and compact by using the regularity condition of the polynomial complementarity problem. Kim et al. DTN proved the existence of Pareto efficient solutions of an unconstrained polynomial vector optimization problem when the Palais-Smale-type condition holds and the image of the objective function has a bounded section. When is a convex semi-algebraic set and is convex, Jiao et al. LGJ proved that has a Pareto efficient solution if and only if the image of has a nonempty bounded section. Inspired by the above works, in this paper, we study the nonemptiness and boundedness of the solution sets of without assuming any convexity of the objective function.
The rest of this paper is organized as follows: In Section 2, we present some notations and preliminary results. In Section 3, we give some sufficient and necessary conditions for regularity conditions. In Section 4, we discuss local properties of the regularity conditions. Section 5 is devoted to the study of solvability of under regularity conditions. In Section 6, we discuss the solution stability under the regularity conditions. Finally, we makes a concluding remark in Section 7.
2 Preliminaries
In this section, we recall some concepts and results that will be used in this paper. A nonempty subset of is called a cone if for any and any . Given a nonempty closed set , the asymptotic cone of is defined by
As known, is a closed cone and , and is bounded if and only if . If is a convex set, then is also a convex cone and , where is the recession cone of defined by
It is known that . The above results can been found in RC ; AA .
Definition 1
Let be a vector polynomial with , . We say that is the vector recession polynomial (or the vector leading term) of , where
Remark 1
When , is a recession polynomial of (see HV1 ).
Definition 2
Definition 3
We say that is weakly regular (resp. strongly regular) if (resp. ) is bounded.
Remark 2
Clearly, strong regularity implies weak regularity. When , both weak regularity and strong regularity coincide with the regularity in (HV1, , Definition 2.1).
The following result plays an important role in establishing the existence of the Pareto efficient solutions of .
Lemma 1
(MG, , Proposition 13) Given and , define and . If , then .
In what follows we always assume the each component polynomial of the objective function has a degree .
3 Regularity of
In this section, we shall discuss properties and characterizations of regularity of .
3.1 Conditions for regularity
In this subsection we shall show that the regularity of is closely related to the regularity of . To do so, we first give a characterization of .
Proposition 1
if and only if .
Proof We only need to prove the sufficiency. Suppose that . Then there exists such that
which yields
Let . It follows that
for all sufficiently large . Since is arbitrary, . ∎
Example 1
Consider the vector polynomial with
and
It is easy to verify that , , and . Then since and . By Proposition 1, .
Proposition 2
If , then is a nonempty cone.
Proof Since , by Proposition 1, . Suppose on the contrary that there exist and such that . Then there exists such that
(1) |
Dividing the both sides of the inequality (1) by , we obtain for all . This reaches a contradiction to since . Thus, is a nonempty cone. ∎
Remark 3
By Proposition 2, PVOP is strongly regular if and only if is empty or .
Proposition 3
If , then is unbounded from below on for all .
Proof Suppose on the contrary that there exists such that is bounded from below on . Then since . By Lemma 1, there exists such that . Since , there exist with and such that as . Since is bounded from below on , there exists a constant such that
Letting , we obtain which is a contradiction to . ∎
When is bounded from below on for some , by Proposition 3 and Proposition 1, we know that . In the following proposition, we further show when each component is bounded from below on .
Proposition 4
If is bounded from below on , then .
Proof Suppose that is bounded from below on . Then is bounded from below on for all . By Proposition 3 and Proposition 1, for all and for all . This means . ∎
The following result shows that the weak (strong) regularity of is closely related to the regularity of .
Theorem 3.1
(i) if and only if for all .
(ii) If , then for all .
Proof (i) Assume that . Since
it suffices to show for all . Suppose on the contrary that there exists such that which is equivalent to (by Lemma 1). Then there exists such that . Let . Consider the function
(2) |
Then is nonempty and closed. Next we shall show that is bounded. If not, then there exists the sequence such that as . Without loss of generality, we assume that and . Since , we have
Dividing the both sides of the above inequality by and letting , we get
which together with yields , a contradiction. Thus, is bounded. By the known Weierstrass’ theorem, we have . It follows from Lemma 1 that
which implies , and so . By the definition of , we get , a contradiction to . Therefore, for all .
For the converse, assume that for all and there exists . Then
which implies that for some . This reaches a contradiction to .
(ii) Suppose on the contrary that there exists such that . By Lemma 1, . Then there exists such that . Let . By similar arguments as in the proof of (i), we have
where and are defined as in (2). The rest is same as the one of (i), and so we omit it. ∎
Remark 4
The following example shows that does not imply for each .
Example 2
Consider the vector polynomial with
and the constraint set
It is easy to verify that , and . Clearly, . However, and .
Proposition 5
is strongly regular and is bounded from below on if and only if for all .
Next we prove the sufficiency. By Theorem 3.1(i), is strongly regular. Suppose on the contrary that there exists such that is not bounded from below on . Let . Then there exists such that
(3) |
for all sufficiently large . We claim that is unbounded. Indeed, if not, then we may assume that and as . Dividing the both sides of (3) by and letting , we get , a contradiction to . Hence, is bounded. Without loss of generality, we may assume that as . It follows from (3) that
a contradiction. Thus, is bounded from below on . ∎
3.2 Regularity and -property
It has been shown in Oett ; FYP ; FangH ; Gowda ; Lop that -property plays an important role in studying the compactness of the solution sets of complementarity problems as well as the upper continuity of the solution mappings. In this subsection, we shall show that the weak (strong) regularity of is closely related to the -property of the following weak vector complementarity problem CGY ; Yangxq
where is the gradient of ,
is the weak dual cone of with respect to , and is the space of all bounded linear operators from to . We denote the solution set of by . Recall that is of type FYP if .
Theorem 3.2
Assume that is convex. If , then is of type . Conversely, if is of type , then is strongly regular.
Proof It is clear that .
Suppose that . By Theorem 3.1(i), we have
This implies for all . This together with the Euler’s Homogeneous Function Theorem yields
for all , where is the degree of . As a consequence, and so is of type .
For the converse, suppose is of type . To complete the proof, it suffices to show (by Remark 3). Suppose on the contrary that there exists . By (LKY, , Theorem 4), we get
This implies since is a closed convex cone. This reaches a contradiction. ∎
The following example illustrates the conclusion of Theorem 3.2.
Example 3
Consider the vector polynomial with and
Then , , and . It is easy to verify that . Let be such that . It follows that This implies . As a consequence, is of type . By Theorem 3.2, is strongly regular.
Remark 5
Example 4
Consider the vector polynomial with and
Clearly, is convex, , and is bounded from below on . It is easy to verify that , , , and . Let . Then and for all . This means , and so is not of type .
4 Local Properties of Regularity Conditions
In this section, we investigate local properties of (strong) weak regularity of . Given an integer , in what follows, we always let denote the family of all polynomials of degree at most , and
whose components are listed by the lexicographic ordering. The dimension of is denoted by . Then, for each polynomial , there exists a unique such that . can be endowed with a norm . Let with and with . It is easy to verify that and as .
Given with being an integer, , let . Denoted by (resp. ) the family of all vector polynomials with , such that is strongly (resp. weakly) regular. Then we have the following results.
Proposition 6
and are nonempty.
Proof We only need to prove that is nonempty since . If is bounded, then . In this case , and so PVOP is strongly regular for every vector polynomial . Suppose that is unbounded. Then there exists . Without loss of generality, we suppose that . Consider the vector polynomial with . Then is a polynomials of degree and as . As a consequence, , and so . ∎
Proposition 7
is open in .
Proof We shall prove that is closed in . Let with such that as . We can suppose that for all since when for some , where is the -th component of . Since is unbounded for all , there exists such that . Without loss of generality, we assume that . We claim that . Indeed, if not, then there exists such that
(4) |
Since , we have
Then for each , there exists such that
Since the set is finite, without loss of generality, we suppose that there exists such that
Since as , dividing the both sides of the above inequality by and letting , we get
This reaches a contradiction to , and so . By Proposition 2, is unbounded. Hence, is closed. ∎
In the following result, we shall show that strong regularity of a polynomial vector optimization problem remains stable under a small perturbation.
Theorem 4.1
The following conclusions hold:
(i) If , then there exists such that for all satisfying ;
(ii) If , then there exists such that for all satisfying .
Proof Since is open in (by Proposition 7) and , there exists an open ball such that either or for all .
(i) It suffices to show that there exists such that for all when . Suppose on the contrary that for any , there exists with such that . By Lemma 1, there exists such that
(5) |
Since as , we have as . Without loss of generality, we assume that as . Since , we get . Dividing the both sides of (5) by and letting , we get
which reaches a contradiction to .
(ii) It suffices to show that there exists such that for all when . Suppose on the contrary that for any , there exists with such that . By (i) of Theorem 3.1, we have
for any . Since as , we have as . Letting in the above inequality, we get
Since is arbitrary, again from (i) of Theorem 3.1 we get , a contradiction. ∎
Observe that for all with , . As a consequence, we have the following result.
Proposition 8
Let be a vector polynomial. Then for any vector polynomial with , the following conclusions hold:
-
(i)
If , then .
-
(ii)
If , then .
-
(iii)
If , then .
-
(iv)
If , then .
Corollary 1
Let be a vector polynomial. Then for any vector polynomial with , the following conclusions hold:
-
(i)
If is weakly regular, then is weakly regular.
-
(ii)
If is strongly regular, then is strongly regular.
5 Existence Results for with Regularity
In this section we shall study emptiness and boundedness of the solution sets of under the regularity condition.
5.1 The weak regularity case
First, we give a necessary condition for the existence of Pareto efficient solutions of .
Theorem 5.1
Assume that one of the following conditions hold:
-
(i)
.
-
(ii)
there exists such that .
Then is nonempty.
Proof Let and . Define and . Clearly, is nonempty and closed. We assert that is bounded. If not, then there exists such that and as . It follows from that
Dividing the both sides of the above inequality by and letting , we get
a contradiction to as well as . Hence, is compact. By the Theorem, . Since (by Lemma 1), we have . ∎
The following example shows that may be unbounded when .
Example 5
Consider the vector polynomial with and
It is easy to see that , , and . Clearly, . On the other hand, is unbounded since
The following result gives a Frank-Wolf type theorem for under the weak regularity condition.
Corollary 2
If is weakly regular and is bounded from below on , then is nonempty.
The following result shows that the existence of the Pareto efficient solutions of a polynomial vector optimization problem is preserved when the vector objective function is perturbed by a lower degree polynomial.
Theorem 5.2
Assume that one of the following conditions hold:
-
(i)
.
-
(ii)
there exists such that .
Then is nonempty for all with , .
5.2 The strong regularity case
In this subsection, we study nonemptiness and boundedness of the solution sets of under the strong regularity condition. Next, we give a necessary condition for the existence of the weak Pareto efficient solutions of .
Theorem 5.3
If , then .
Proof Suppose that there exists . By Proposition 1, . Then there exists such that
This means that
(6) |
Since , there exist with and such that as . Since and , there exists such that
Without loss of generality, we can assume that there exists such that
Dividing the both sides of the above inequality by and letting , we have
a contradiction to (6). ∎
Remark 7
By Theorem 5.3, implies that . The following example shows that the converse does not hold in general.
Example 6
Consider the vector polynomial with
and . It is easy to verify that . On the other hand, and , but as . This implies .
Theorem 5.4
If , then .
Remark 8
Example 7
Consider the vector polynomial with
and
Then and . It is easy to verify that and .
The following result shows that is sufficient for the existence of the Pareto efficient solutions as well as boundedness of the weak Pareto efficient solution set.
Theorem 5.5
If , then is nonempty and is compact.
Proof follows from Theorem 3.1(i) and Theorem 5.1(ii). The closedness of is clear. Next we prove that is bounded. If not, then there exists such that as . Without loss of generality, assume that and . Let . Since , there exists such that
Since the set is finite, without loss of generality, we can assume that there exists such that
Dividing the both sides of the above inequality by and letting , we have . This implies , a contradiction. ∎
Corollary 3
If for each , then is nonempty and is bounded.
Now we give an example to illustrate the conclusion of Corollary 3.
Example 8
Remark 9
Example 9
Consider the vector polynomial with
and
It is easy to verify that , , , . On the other hand, is unbounded since
It has been shown in Corollary 2 that with weak regularity admits a weak Pareto efficient solution provided that is bounded from below on . In the following Frank-Wolf type theorem, we further prove the existence of Pareto efficient solutions and compactness of the weak Pareto efficient solution set if we strengthen weak regularity by strong regularity.
Corollary 4
If is strongly regular and is bounded from below on , then is nonempty and is compact.
The following example shows that the converse of Theorem 5.5, Corollary 3 and Corollary 4 does not hold in general.
Example 10
Consider the polynomial with
and
Then and . It is easy to verify that . However, are unbounded.
The following result shows that some properties of the solution sets of a strongly regular polynomial vector optimization problem are preserved when its objective polynomial is perturbed by a lower degree polynomial.
Theorem 5.6
Let be a vector polynomial with . Then the following conclusions hold:
-
(i)
If , then .
-
(ii)
If , then is nonempty and is compact.
6 Stability Analysis
In this section we investigate the solution stability of a regular polynomial vector optimization problem. First, we shows that some properties of the solution sets of a strongly regular polynomial vector optimization problem are stable under a small perturbation.
Theorem 6.1
The following conclusions hold:
-
(i)
If , then there exists such that for all satisfying .
-
(ii)
If , then there exists such that is nonempty and is compact for all satisfying .
Proof (i) follows from Theorem 4.1 and Theorem 5.3 and (ii) follows from Theorem 4.1 and Theorem 5.5. ∎
In the sequel we shall investigate the local boundedness and upper semicontinuity of the weak Pareto efficient solution mapping. Recall that a set-valued mapping is said to be upper semi-continuous at if for any neighborhood of , there exists a neighborhood of such that for any . A set-valued mapping is locally bounded at if there exists an open neighborhood of such that is bounded.
Lemma 2
(RC, , Theorem 5.7 and Theorem 5.19) If is locally bounded at and has a closed graph , where , then is upper semi-continuous at .
Theorem 6.2
Assume that is convex and is strongly regular. Then the following conclusions hold:
-
(i)
is locally bounded at , i.e., there exists such that
is bounded, where the is an open ball in with center at and radius .
-
(ii)
is upper semi-continuous on .
Proof (i) Since is open in (by Proposition 7) and , there exists such that , where denotes the closure of . Suppose on the contrary that is unbounded. Then there exist and such that as . Without loss of generality, we may assume that and . Let and . Then for all . Since , we have
Then for each , there exists such that
Since is finite, we can assume that there exists such that
Dividing the both sides of the above inequality by and letting , we get
Since is arbitrary, we have . On the other hand, since , by Proposition 2 we have , a contradiction. Thus, is locally bounded at .
(ii) Let . By (i) and Lemma 2, we only need to prove that the graph of is closed in . Let as with . Then for any , we have
Letting , we have
which means . Thus, is closed in .. ∎
7 Conclusion
In this paper we extend the concept of regularity due to Hieu HV1 to the polynomial vector optimization problem. Under regularity conditions, we investigate nonemptiness and boundedness of the solution sets of a non-convex polynomial vector optimization problem on a nonempty closed set (not necessarily semi-algebraic set). As a consequence, we derive two Frank-Wolfe type theorems for a non-convex polynomial vector optimization problem. We also discuss the solution stability. Our results extend and improve the corresponding results of DTN ; BT1 ; BT2 ; HV1 ; LGJ .
References
- (1) Ha, H.V., Pham, T.S.(eds.): Genericity in Polynomial Optimization. World Science Publishing, Singapore (2017)
- (2) Benedetti, R., Risler, J.J.(eds.): Real Algebraic and Semi-Algebraic Sets. Hermann, Paris (1990)
- (3) Ansari, Q.H., Lalitha, C.S., Mehta, M.(eds.): Generalized Convexity, Nonsmooth Variational Inequalities, and Nonsmooth Optimization. Chapman and Hall/CRC, New York (2013)
- (4) Hieu, V.T.: A regularity condition in polynomial optimization. arXiv:1808.06100
- (5) Hieu, V.T., Wei, Y.M., Yao, J.C.: Notes on the optimization problems corresponding to polynomial complementarity problems. J. Optim. Theory Appl. 184, 687–695 (2020)
- (6) Kim, D.S., Pham, T.S., Tuyen, V.T.: On the existence of Pareto solutions for polynomial vector optimization problems. Math. Program. 177, 321–341 (2019)
- (7) Jiao, L.G., Lee, J.H., Sisarat, N.: Multi-objective convex polynomial optimization and semidefinite programming relaxations. arXiv:1903.10137
- (8) Rockafellar, R.T., Wets, R.J-B.(eds.): Variational Analysis. Springer, Berlin (2009)
- (9) Auslender, A., Teboulle, M.(eds.): Asymptotic Cones and Functions in Optimization and Variational Inequalities. Springer, New York (2003)
- (10) Bao, B.T., Mordukhovich, B.S.: Variational principles for set-valued mappings with applications to multiobjective optimization. Control Cybern. 36, 531–562 (2007)
- (11) Bao, B.T., Mordukhovich, B.S.: Relative Pareto minimizers for multiobjective problems: existence and optimality conditions. Math. Program. 122, 301–347 (2010)
- (12) Giannessi, F., Mastroeni, G., Pellegrini, L.: On the theory of vector optimization and variational inequalities. Image space analysis and separation. In: Giannessi, F. (ed.): Vector Variational Inequalities and Vector Equilibria. Nonconvex Optimization and Its Applications, vol. 38, Springer, Boston, MA. (2000)
- (13) Huang, N.J. and Fang, Y.P.: The upper semicontinuity of the solution maps in vector implicit quasicomplementarity problems of type . Appl. Math. Lett. 16, 1151–1156 (2003)
- (14) Fang, Y.P. and Huang, N.J.: On the upper semi-continuity of the solution map to the vertical implicit homogeneous complementarity problem of type . Positivity 10, 95–104 (2006)
- (15) Oettli, W., Yen, N.D.: Quasicomplementarity problems of type , J. Optim. Theory Appl. 89, 467–474 (1996)
- (16) R. López: Stability results for polyhedral complementarity problems. Comput. Math. Appl. 58. 1475–1486 (2009)
- (17) Gowda, M. S., Sossa, D.: Weakly homogeneous variational inequalities and solvability of nonlinear equations over cones. Math. Program., Ser. A, 177, 149–171 (2019)
- (18) Yang, X. Q.: Vector complementarity and minimal element problems. J. Optim. Theory Appl. 77, 483–495 (1993)
- (19) Chen, G.Y., Yang, X.Q.: The vector complementary problem and its equivalence with weak minimal elements in ordered spaces. J. Math. Anal. Appl. 153, 136–158 (1990)
- (20) Lee, G.M., Kim, D.S., Lee, B.S., Yen N.D.: Vector variational inequality as a tool for studying vector optimization problems. In: Giannessi, F.(ed.) Vector Variational Inequalities and Vector Equilibria. Nonconvex Optimization and Its Applications, vol. 38, Springer, Boston, MA. (2000)