Talk:Subgradient method

I do not understand, are the convergence proofs for all 5 step size choices known? I think this contradicts Nesterov[07] paper where it is claimed that the convergent variants of this method have step sizes that satisfy the divergent series rule - sum of steps=infinity, but limit of step =0.


 * Welcome to Wikipedia!
 * Bertsekas has a rather comprehensive study of step-size rules. (Please remember to sign your posts.) Kiefer .Wolfowitz 18:15, 9 May 2011 (UTC)