Talk:Sequential minimal optimization

Robotics Unassessed

	This article is within the scope of WikiProject Robotics, a collaborative effort to improve the coverage of Robotics on Wikipedia. If you would like to participate, please visit the project page, where you can join the discussion and see a list of open tasks.RoboticsWikipedia:WikiProject RoboticsTemplate:WikiProject RoboticsRobotics
???	This article has not yet received a rating on Wikipedia's content assessment scale.
???	This article has not yet received a rating on the project's importance scale.

Delete this article?

This article lacks references to published reliable papers. Moreover it doesn't coherently describe anything notable or of interest: The smallest possible QP is the zero QP on the zero vector, which has a constant extension everywhere. Would somebody nominate this for deletion? Thanks Kiefer.Wolfowitz (talk) 13:03, 8 November 2010 (UTC)[reply]

Hold on, I'll to try the fill the article with some information by tomorrow. The subject of the article is certainly notable - there are many articles and books which describe SMO, e.g. [1], [2]. The reference currently cited is Platt's original paper, and it's well-cited (819 citations on Google Scholar, 165 on CiteSeer), so it's reliable. -- X7q (talk) 13:34, 8 November 2010 (UTC)[reply]

SMO is an working-set method - at each iteration it optimizes a subset of variables and leaves the rest fixed. "Smallest possible QP problem" means that it chooses the smallest possible number of variables to optimize at each step, but which still allows it to make a progress. And this number is 2, because SVM optimization problem in dual-form has one equality constraint. With no constraints it could optimize one variable (i.e. it'd be basically coordinate descent), but due to an equality contraint it's not allowed to change value of any variable without changing value of some other variable. -- X7q (talk) 13:34, 8 November 2010 (UTC)[reply]

But I think SMO should be removed from Template:Optimization_algorithms. It's an optimization method for a very specific problem, not a general-purpose QP method. -- X7q (talk) 13:41, 8 November 2010 (UTC)[reply]

Thanks for a quick and helpful response. It sounds like this might be a notable topic, but the article is still a little rough. I think that your suggested improvements should be very helpful. Thanks again, Kiefer.Wolfowitz (talk) 13:47, 8 November 2010 (UTC)[reply]