solve a maximize problem by YALMIP

160 views
Skip to first unread message

reza mahjoub

unread,
Sep 2, 2014, 11:13:01 AM9/2/14
to yal...@googlegroups.com
max D'AD
subject to: d'd<= E
 
D is m*n and A is m*m. 
E=constant
does it convex?and solvable by YALMIP?

reza mahjoub

unread,
Sep 2, 2014, 11:20:01 AM9/2/14
to yal...@googlegroups.com
the goal is finding optimum D.

reza mahjoub

unread,
Sep 2, 2014, 11:21:22 AM9/2/14
to yal...@googlegroups.com
the goal is finding optimum D. "d" is "D". I wrote it wrong.

Johan Löfberg

unread,
Sep 2, 2014, 12:17:51 PM9/2/14
to
No it is not convex.

Yes it is "solvable" by YALMIP

However, it is trivially solved by computing eigenvalues (D will by definition be a scaled eigenvector associated with the largest eigenvalue of A). I assume you have a typo and D is a vector, as it makes no sense to maximize a matrix D'*A*D.

reza mahjoub

unread,
Sep 2, 2014, 12:48:10 PM9/2/14
to yal...@googlegroups.com
Thank you so much.
 
Yes. sorry D is a vector. If its not convex, then what type is that? 

Johan Löfberg

unread,
Sep 2, 2014, 1:36:17 PM9/2/14
to
Depends on A. If A is positive semidefinite, you are maximizing a convex function, i.e., minimizing a concave function

Hence, if A is negative definite, you have a convex problem as you would be maximizing a concave function, i.e., minimizing a convex function.
Message has been deleted

reza mahjoub

unread,
Sep 2, 2014, 3:29:49 PM9/2/14
to yal...@googlegroups.com
Thanks a lot.
Reply all
Reply to author
Forward
0 new messages