In this blog, I hope to collect and discuss a variety of the more common questions, issues, and topics I encounter as I develop and support CVX and TFOCS, and as I teach and speak about convex optimization. Most of these posts will be written by me, Michael Grant. But perhaps with some prodding, we’ll be able to convince certain heavyweights in the field to offer their thoughts here, too.
Over time I hope to illuminate and clarify the notion of disciplined convex programming. It is not just a class of optimization problems that by chance have attractive geometry. Rather, it is a specific methodology for building optimization models. Consider linear programming: people intending to use an LP tool know that nonlinear functions are not accepted. This knowledge shapes their thinking as they build models. Arguably, it limits their choices, but it also encourages them to think harder about how they can express their models effectively without nonlinearity. The result? Linear programming has proven to be an extremely successful methodology with application in a wide variety of fields.
So it is with disciplined convex programming, only our restrictions have actually been relaxed: nonlinearities can now be considered, as long as they are convex. In theory, a much wider variety of problems can be effectively modeled this way (including, of course, every linear program). So as the tools mature I am certain that convex optimization will grow in adoption.
Speaking of those tools, this blog will also provide tips and techniques on how to make better use of my packages, CVX and TFOCS. There are a number of other excellent convex optimization packages out there as well, and I have links to several of them on the left sidebar. Many of the tips I share here will apply to those packages as well.
So watch this space! Add this site to your bookmarks; and please feel free to offer comments and suggestions on topics to discuss.
EDIT: It seems that I managed to release this blog to the search engines before I was ready. Thanks for the initial comments, but I’m not quite there yet. I will get this going soon.
Thanks for visting,
Michael