Global questions about cvx primal-dual solver


(Mykola Servetnyk) #1

Dear CVX users,

I have very general question to you. I know how to obtain dual problems for some standard kind of programming(LP, QCQP and so on). CVX in return knows how to obtain dual problems for any complicated formulations, that just satisfy DCP ruleset. Assume that I have a convex problem, satisfying DCP rules. As I know CVX can solve dual of that problem, even though it is not in standard form. My question is how CVX gets the dual of the problem? Is there any place to read about its powerful mechanism obtaining the dual? I want to know, how it ‘turns the crank’. I have doubts that it composes the Lagrangian function, takes derivatives and so far and so on.

Thanks!


(Mark L. Stone) #2

Read mcg’s thesis “Disciplined Convex Programming” https://stanford.edu/~boyd/papers/pdf/mcg_thesis.pdf .

If you have any questions on his thesis, or want to find out about subsequent developments in CVX’s under the hood processing, he’s the right one to answer.