Proximal Methods for Cohypomonotone Operators

From MaRDI portal





The paper under review is a valuable contribution to nonlinear functional analysis and, in particular, algorithms for finding a zero of one (or several) operator(s) \(A\) in a Hilbert space. While the first approach is given by step-wise matrix multiplication with the inverse operator of \(\text{Id} + A\), more refined versions are presented by the proximal point iteration with a ``small additive term per iterate included, by the relaxed version of this one and, finally, by replacing \(A\) with a member of a countable family of operators \(A_i\) which may even be set-valued. Hence, the zero problem is of inclusion form. The author's first algorithm refers to such a generalized version, where the operators are assumed to be maximally monotone and bounds are given on the sequences of parameters, e.g., to the model errors. Then, every such orbit converges weakly to a point in the common zero set of the operators. Further important results of such a convergence type are obtained by studying special classes of operators. In fact, there are (\(\rho\)-)cohypomonotone operators which become well-introduced and investigated. The corresponding proximal iteration theorem with its special assumptions, e.g., concerning the parameter \(\rho\), are carefully stated and proved. The paper ends with an application of the author's approach to nonlinear programming (in Euclidean space) with finitely many equality and inequality constraints. Here, the algorithm is well-related with the optimization terminology, e.g., a KKT system in inclusion form and a strong second sufficient order condition is provided in order to conclude convergence.




Cited in
(36)








This page was built for publication: Proximal Methods for Cohypomonotone Operators

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4652557)