A Proximal Bundle Method Based on Approximate Subgradients

摘要: In this paper a proximal bundle method is introduced that capable to deal with approximate subgradients. No further knowledge of the approximation quality (like explicit or controllability error bounds) required for proving convergence. It shown every accumulation point sequence iterates generated by proposed algorithm well-defined solution exact minimization problem. case subgradients behaves like well-established methods. Numerical tests emphasize theoretical findings.