Studies on common solutions of a variational inequality and a fixed-point problem (Q1937090): Difference between revisions

The authors consider a problem of finding an element of the set \(\Omega_1\cap\Omega_2\) where \[ \Omega_1=\{ u \in K \;| \;\langle Tu, v-u\rangle \geq 0 \;\forall v \in K \}, \;\Omega_2=\{ u \in K \;| \;Su=u\}, \] \(K\) is a convex and closed set in a real Hilbert space \(H\), \(T: K \to H\) is an inverse strongly monotone (co-coercive) mapping, and \(S: K \to K\) is a strictly pseudo-contractive mapping (i.e., \(I-S\) is co-coercive). They suggest an alternating projection step method and claim that it provides a strong convergence if \(\Omega_1\cap\Omega_2 \neq \emptyset\) (Theorem 3.2). However, this assertion is incorrect. In fact, by setting \(S\equiv I\), we reduce the proposed method to the usual projection one, which converges only weakly to a point of \(\Omega_1\), i.e. to a solution of the variational inequality problem with co-coercive mapping. The mistake is contained in the proof of Claim 2 (p. 532--533), because the bounded sequence \(\{ x_n \}\) has only weakly (not strongly) convergent subsequences and the auxiliary mappings \(R_T\) and \(R_T\) are continuous, but not weakly continuous. Thus, the assertion is true only in a finite-dimensional setting.