Yalmip:SDP problem, run into numerical problem. Is the result credible, when the program run into numerical problem?

Question

QING ZHOU 2021-3-23

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/780722-yalmip-sdp-problem-run-into-numerical-problem-is-the-result-credible-when-the-program-run-into-nu

评论： Johan Löfberg 2021-3-31

The 0.9999 is expected result. However, there may be something wrong with the procedure, as it shows: run into numerical problem. Is the result credible, when the program run into numerical problem?

In other word, if a numerical problem has occurred, can the result of this program be considered to be exact ?

Thanks in advance!

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Johan Löfberg 2021-3-31

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/780722-yalmip-sdp-problem-run-into-numerical-problem-is-the-result-credible-when-the-program-run-into-nu#answer_663289

You should post YALMIP specific questions on the YALMIP forum.

Short answer it depends. You will have to study the solution and see if it satisfies your constraints, and if it is good enough. The duality gap is on the order 10^-7 which is sort of ok, but it is struggling with feaibility so the gap measure could be questioned (read sedumi documentation to interpret the log)

2 个评论
显示无隐藏无

QING ZHOU 2021-3-31

Thanks for your answer.

If i change the value of pars.eps in sedumi documentation as 5e-7, 'run into numerical problem' will not appear.

Johan Löfberg 2021-3-31

Correct, if you ask he solver not to tell you that it has numerical problems, it will not tell you that.

I use the same strategy when kids tell me they are sick. I ask them to not convey that information to me. I don't know if it actually changes anything regarding their health state though...

请先登录，再进行评论。