Hi,
I am using matlab command "quadprog" to solve a QP problem for SVM regression.
But quite often, I get a warning that the problem is badly conditioned and so
the solution may be unreliable. Infact in such cases the solution to QP problem
yields a result, that is way off from the expected result. Could anyone tell me
what exactly bad conditioned problem means, and also if possible suggest ways
to avoid it.
Thanks in advance.
--
Amaresh V Malipatil
Graduate Student
Department of Electrical Engineering
University of Notre Dame
Indiana 46556
|