Ok, I've already asked this question in math.stackexchange, but I feel it is more appropriate to ask here (hopefully I am not violating any rules by repeating!). So here it is:

I wonder if there is a local convergence proof for ADMM applied to biconvex problems? More specifically, my problem is as follows:

$$\min_{x,y} f(x) + g(y) + \| y \circ Ax \|_2^2 $$ ,

where $\circ$ denotes Hadamard product and $f(x)$ and $g(y)$ are convex functions. This fits into biconvex setting mentioned in Boyd's paper on distributed ADMM, but without theoretical evidence.

Thank you!


To partially answer my question - there is this paper, but coming with a strong assumption on the multiplier value. Mingyi Hong, Zhi-Quan Luo, Meisam Razaviyayn "Convergence Analysis of Alternating Direction Method of Multipliers for a Family of Nonconvex Problems"

  • $\begingroup$ Please provide the full citation (since links are not usually permanent). $\endgroup$ – Paul Oct 29 '15 at 1:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.