Convergence of a generalized subgradient method for nondifferentiable convex optimization (Q757242): Difference between revisions

The generalized subgradient method proposed in this paper for minimizing a convex function f: \(R^ n\to R\) generates a sequence of points by an iteration of the form \[ x_{k+1}=x_ k-\sum_{i\in I_ k}s^ i_ kg_ i, \] where \(I_ k\subset \{1,...,k\}\), \(g_ i\in \partial f(x_ i)\) and \(s^ i_ k\geq 0\). Observing that this iteration can be written as \(x_{k+1}=x_ k-s_ kd_ k\), with appropriate \(s_ k\geq 0\) and \(d_ k\in \partial_{\epsilon_ k}f(x_ k)\) for some \(\epsilon_ k\geq 0\), the authors derive some convergence conditions from known convergence theorems for the \(\epsilon\)-subgradient method.