Abstract
A new interior proximal method for variational inequalities with generalized monotone operators is developed. It transforms a given variational inequality (which, maybe, is constrained and ill-posed) into unconstrained and well-posed equations as well as, at each iteration, one single additional extragradient step with rather small numerical efforts. Convergence is established under mild assumptions: The frequently assumed maximal monotonicity is weakened to pseudo- and quasimonotonicity with respect to the solution set, and a wide class of even nonlinearly constrained feasible sets is allowed for. In this general setting, the presented scheme constitutes the first interior proximal method that works without the so-called cutting plane property. Such a demanding assumption is completely left out, which allows to solve, e.g., wide classes of saddle point and equilibrium problems by means of an interior proximal method for the first time. As another application, we study variational inequalities derived from quasiconvex optimization problems.
Similar content being viewed by others
Notes
Notions of monotonicity are introduced below.
The reason is that in the sequel, we will work with multiple subsequences, i.e., a subsequence of a subsequence of..., and to this end, a notation as \(x^{k_{l_{m_{\ldots }}}}\) is rather complicated.
Even if \(T\) is maximal monotone, our analysis provides new results. For instance, the feasible set may significantly more general than \(K=\mathbb {R}^n_+\), as in [5].
It might be of interest that (7) can also be equivalently written as a variational inequality (see [16, Lemma 3.5]): Find \(y^k \in \mathbb {R}^n\) such that, for some error parameter \(\delta _k \ge \Vert e^k\Vert \),
$$\begin{aligned} \langle t^k(y^k) + \chi _k \left( \nabla h(y^k) - \nabla h(x^k)\right) , x-y^k\rangle \ge -\delta _k \Vert x-y^k\Vert , \quad \forall ~ x \in K. \end{aligned}$$(8)It is shown later that our assumptions indeed ensure that \(\{x^k\} \subset \mathrm {int}K\).
Burachik and Dutta worked with \(\vartheta (x) = c_1 x\) for some \(c_1 > 0\). But it might also be interesting to work with, e.g., \(\vartheta (x) = \sqrt{x}\). However, apart from that slight difference, the error criterion is exactly the same as in [5].
Necessary changes for quasimonotonicity are discussed once convergence is established for the pseudomonotone case.
The case \(\rho = + \infty \) is not excluded; we comment on this soon. Anyway, the subsequent analysis works regardless of whether \(\rho < + \infty \) or not.
If \(T\) is monotone, we identify \(L = 0\) and define \(L/\kappa :=0\), regardless of whether \(\kappa = 0\).
Loosely speaking, in view of Theorem 4.1 below this indicates that this function \(\varphi \) should, in general, only be applied if \(K\) is polyhedral.
We mention that, for \(K=\mathbb {R}^n_+\), we recover the assumption \(r_k \le \min _{i=1,\ldots ,n} x_i\) as worked with in [5].
Following a combination of the references [9, 23, 28], other examples may also be deduced, e.g., for \(K\) being a ball. Of course, in case of more complex and / or unbounded feasible sets \(K\) it is rather unlikely that \(\nabla h\) is explicitly invertible. Anyway, the extragradient step then still appears to be numerically attractive due to the remaining properties of \(\nabla h\), e.g., strict monotonicity and zone coerciveness.
Note that in the proof of Theorem 5.2 no special property of \(T\) (in particular, no cutting plane property nor any related assumption) is required.
In combination with Lemma 4.1, this establishes what has been announced above: If \(\overline{x} \in \mathrm {int}K\), and regardless of whether \(e^k \rightarrow 0\), then by Lemma 4.1, \(g_j(\overline{x}) < 0\) for all \(j=1,\ldots ,m\). Hence, \(I(\overline{x}) = \emptyset \) and the sum of the inferior limits is zero. In that case, no further argumentation is necessary. It is therefore enough to consider \(i \in I(\overline{x})\) only.
The remaining \(i \in I(\overline{x})\) will be dealt with subsequently in an iterative manner.
References
Martinet, B.: Régularisation d’inéquations variationelles par approximations successives. Revue française d’informatique et de recherche opérationnelle, série rouge 4, 154–158 (1970)
Rockafellar, R.T.: Monotone operators and the Proximal Point Algorithm. SIAM J. Control Optim. 14, 877–898 (1976)
Auslender, A., Teboulle, M., Ben-Tiba, S.: A logarithmic-quadratic proximal method for variational inequalities. Comput. Optim. Appl. 12(1–3), 31–40 (1999)
Da Silva, P.J., Eckstein, J., Humes, C.: Rescaling and stepsize selection in proximal methods using separable generalized distances. SIAM J. Optim. 12, 238–261 (2001)
Burachik, R., Dutta, J.: Inexact proximal point methods for variational inequality problems. SIAM J. Optim. 20, 2653–2678 (2010)
Teboulle, M.: Convergence of proximal-like algorithms. SIAM J. Optim. 7, 1069–1083 (1997)
Kaplan, A., Tichatschke, R.: Interior proximal method for variational inequalities: case of non-paramonotone operators. J. Set-valued Anal. 12, 357–382 (2004)
Burachik, R., Iusem, A.: A generalized Proximal Point Algorithm for the variational inequality problem in Hilbert space. SIAM J. Optim. 8, 197–216 (1998)
Censor, Y., Iusem, A., Zenios, S.: An interior point method with Bregman functions for the variational inequality problem with paramonotone operators. Math. Program. 81, 373–400 (1998)
Censor, Y., Zenios, S.: Proximal minimization algorithm with D-functions. J. Optim. Theory Appl. 73, 451–464 (1992)
Eckstein, J.: Approximate iterations in Bregman-function-based proximal algorithms. Math. Program. 83, 113–123 (1998)
Kaplan, A., Tichatschke, R.: Bregman functions and proximal methods for variational problems with nonlinear constraints. Optimization 56, 253–265 (2007)
Langenberg, N.: Convergence analysis of an extended Auxiliary Problem Principle with various stopping criteria. Optim. Method Softw. 26, 127–154 (2011)
Solodov, M., Svaiter, B.: An inexact hybrid generalized Proximal Point Algorithm and some new results on the theory of Bregman functions. Math. Oper. Res. 25, 214–230 (2000)
Langenberg, N.: Pseudomonotone operators and the Bregman Proximal Point Algorithm. J. Glob. Optim. 47, 537–555 (2010)
Langenberg, N.: An interior proximal method for a class of quasimonotone variational inequalities. J. Optim. Theory Appl. 155, 902–922 (2012)
Langenberg, N.: On the cutting plane property and the Bregman Proximal Point Algorithm. J. Conv. Anal. 18, 601–619 (2011)
Burachik, R.S., Iusem, A.N.: Set-valued mappings and enlargements of monotone operators. Springer, New York (2008)
Facchinei, F., Pang, J.S.: Finite-dimensional variational inequalities and complementarity problems. Springer, New York (2003)
Chen, G., Teboulle, M.: Convergence analysis of a proximal-like minimization algorithm using Bregman functions. SIAM J. Optim. 3, 538–543 (1993)
Bauschke, H., Borwein, J.: Legendre functions and the method of random Bregman projections. J. Conv. Anal. 4, 27–67 (1997)
Bregman, L.: The relaxation method of finding the common points of convex sets and its application to the solution of problems in convex programming. USSR Comput. Math. Math. Phys. 7, 200–217 (1967)
Kaplan, A., Tichatschke, R.: On inexact generalized proximal methods with a weakened error tolerance criterion. Optimization 53, 3–17 (2004)
Da Silva, P.J., Eckstein, J.: Double-regularization proximal methods with complementarity applications. Comput. Optim. Appl. 33, 115–156 (2006)
Eckstein, J., da Silva, P.J.: Proximal methods for nonlinear programming: double regularization and inexact subproblems. Comput. Optim. Appl. 46, 279–304 (2010)
Crouzeix, J.-P.: Pseudo-monotone variational inequality problems: existence of solutions. Math. Prog. 78, 305–314 (1997)
Kien, B.T., Yao, J.-C., Yen, N.D.: On the solution existence of pseudomonotone variational inequalities. J. Glob. Optim. 41, 135–145 (2008)
Langenberg, N.: Applications to the Nash Equilibrium Problem. Convergence analysis of proximal-like methods for variational inequalities and fixed point problems. Logos Verlag, Berlin (2011)
Kaplan, A., Tichatschke, R.: Note on the paper: interior proximal method for variational inequalities on non-polyhedral sets. Disc. Math.: Diff. Incl. Control Optim. 30, 51–59 (2010)
Aussel, D., Hadjisavvas, N.: Technical note: On quasimonotone variational inequalities. J. Optim. Theory Appl. 121, 445–450 (2004)
Hadjisavvas, N., Schaible, S.: Quasimonotone variational inequalities in Banach spaces. J. Optim. Theory Appl. 90, 95–111 (1996)
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Langenberg, N. Interior Proximal Method Without the Cutting Plane Property. J Optim Theory Appl 166, 529–557 (2015). https://doi.org/10.1007/s10957-014-0605-8
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10957-014-0605-8
Keywords
- Variational inequalities
- Bregman distances
- Proximal Point Algorithm
- Extragradient step
- Generalized monotonicity
- Interior point effect