On approximation of a regularized solution to a nonlinear equation (Q1360842)

The article is devoted to studying the operator equation \(Au=f\), where \(A\) is a nonlinear operator. Suppose that the right-hand side \(f\) is given with some error; i.e., we know \(f_\delta \) rather than \(f\). The \textit{regularized solution} \(M_\delta ^\alpha \) to \(Au=f\) with \(f=f_\delta \) is the set of solutions to the variational problem \(\inf\{|Au-f_\delta |^{p_1}+\alpha |u|^{p_2}\mid u\in \mathop{\fam 0 dom} A\}\), with \(\alpha >0\), \(p_1,p_2\geq 1\). Let \(\{A_n\}\) be a sequence of operators which approximate \(A\). The \textit{approximation} of the regularized solution \(M_\delta ^\alpha\) is the set of solutions to the corresponding variational problem with \(A=A_n\). Sufficient conditions are given for \(\beta \)-convergence of approximations of a regularized solution and for \(\beta \)-convergence of regularized solutions. To this end, the author introduces the notion of an upper semiclosed pair. Moreover, there is an example which demonstrates that, unlike the linear case (see, e.g., the article by \textit{V. P. Tanana} and \textit{A. R. Danilin} [Sov. Math., Dokl. 25, 841-842 (1982); translation from Dokl. Akad. Nauk SSSR 264, 1094-1096 (1982; Zbl 0535.41039)], the weak closure condition for the pair \(A,\;\{A_n\}\) is not necessary for convergence of finite-dimensional approximations.