Generalized Bundle Methods

From MaRDI portal





The author considers the following problem: \[ \inf\bigl\{f(x):x\in X \bigr\}, \tag{P} \] where \(f\) is a convex real-valued function defined on \(\mathbb{R}^n\), \(X\subset\mathbb{R}^n\) is a closed convex set. Basing on the idea of using the boundle \(\beta\) for constructing a model \(f\) of the original function \(f\), he proposes a number of generalized boundle algorithms for solving problem (P) and studies rules for the \(\beta\)-strategy and the \(t\)-strategy. He proves finite termination, asymptotic convergence and finite convergence to an optimal point, with or without limits on the number of serious steps and requiring the proximal parameter to go to infinity. Convergence proofs of several variants of algorithms are given, which contain the convergence of null step sequences, the convergence of serious step sequences, and the ``third level that is necessary for some classes of stabilizing terms. Some extensions of generalized boundle methods, which involve constrained optimization and decomposable functions are given, and relationships with other algorithms from the literature are analyzed as well.




Cited in
(89)








This page was built for publication: Generalized Bundle Methods

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4785872)