Auxiliary problem principle and proximal point methods (Q5928209)

In the present paper the auxiliary problem principle (APP) of \textit{G. Cohen} [J. Optimization Theory Appl. 59, No. 2, 325-333 (1988; Zbl 0628.90066)] is studied for variational inequalities of the type of finding \(x^* \in K: \langle {\mathcal F}(x^*)+ {\mathcal Q} (x^*), x-x^*\rangle \geq 0\) \(\forall x \in K\), where K is a convex closed subset of a Hilbert space X, \(\mathcal F\) is a single-valued operator from X into the dual space \(X^{'}\) and \(\mathcal Q : X \rightarrow 2^{X^{'}}\) is a maximal monotone (multi-valued) operator. The current auxiliary problem is constructed by fixing \(\mathcal F\) at the previous iterate, whereas \(\mathcal Q\) or its single-valued approximation \({\mathcal Q}^k\) is considered at a variable point. Using auxiliary operators of the form \( { \mathcal L}^k + {\chi}_k \nabla h\), with a scalar \( {\chi}_k >0\), a monotone operator \({\mathcal L}^k : X \rightarrow X^{'}\) and an auxiliary functional \( h: X \rightarrow R\) being convex Gateaux-differentiable, the standard for the auxiliary problem principle assumption of the strong convexity of \( h\) can be weakened exploiting mutual properties of \(\mathcal Q\) and \(h\): the function \(h\) is supposed to be convex and the operators \( { \mathcal L}^k + {\chi}_k \nabla h\) have to be strongly monotone with a common modulus for all k. The scheme is referred to as the proximal auxiliary problem (PAP) method, its assumptions are described in Section 2, its convergence (Theorems 1-2) is analyzed in Section 3 together with respective assumptions and three preliminary lemmas. The proof of Lemma 2 is devoted to the appendix. In the final section the applications of the PAP method to different types of variational inequalities are sketched on the basis of approaches of decomposition, linear approximation and weak regularization.