The problem of maximizing a concave nondifferentiable functional on a closed convex subset of a Banach space is considered. The authors propose a new algorithm which yields not only the argument of the maximum in the limit, but also that subgradient involved in the necessary and sufficient optimality conditions. The relevance of this algorithm is shown for computing a saddle point of a convex-concave Lagrangian-like functional when the argument of the inner minimum in the maximum approach is nonunique.
|Number of pages||9|
|State||Published - 1981|
|Event||Proceedings Eighteenth Annual Allerton Conference on Communication, Control, and Computing - Monticello, United States|
Duration: 8 Oct 1980 → 11 Oct 1980
|Conference||Proceedings Eighteenth Annual Allerton Conference on Communication, Control, and Computing|
|Period||8/10/80 → 11/10/80|