Revision as of 15:35, 9 October 2010 edit Kiefer.Wolfowitz (talk \| contribs) 39,688 edits m Adding category Category:Convex optimization (using HotCat) ← Previous edit		Revision as of 19:10, 13 October 2010 edit undo X7q (talk \| contribs) Rollbackers 1,695 edits →Convergence results: It doesn't always converge with constant step size! E.g. try to minimize abs(x) starting from 0.5 with step 1: it oscillates. Next edit →
Line 27: ===Convergence results=== For ~~constant step size and~~ constant step length, the subgradient algorithm is guaranteed to converge to within some range of the optimal value, i.e.,{{cn}} :<math>\lim_{k\to\infty} f_{\rm{best}}^{(k)} - f^* <\epsilon.</math>

Subgradient method: Difference between revisions