What function can I use for finding global optima in semi-infinite constraint optimization problem?

Question

siddhesh rane 2017-10-22

1
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/362625-what-function-can-i-use-for-finding-global-optima-in-semi-infinite-constraint-optimization-problem

评论： siddhesh rane 2017-10-22

采纳的回答： Matt J

在 MATLAB Online 中打开

Below is the simplified version of problem I am trying to solve:

Minimize f(x) =l1+l2+l3

Subject to

     20°≤θ1≤160°
     200°≤θ2≤340°
     200°≤θ3≤340°

Where, θ1,θ2,θ3=f(l1,l2,l3,ϕ)

30°≤ϕ≤150°

I believe this problem falls under semi-infinite constraints problem, how can I find out global optima of it? I read about fseminf function but it only gives local minima. Can it be optimized using GA?

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

Matt J 2017-10-22

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/362625-what-function-can-i-use-for-finding-global-optima-in-semi-infinite-constraint-optimization-problem#answer_287112

编辑：Matt J 2017-10-22

I think a viable strategy would be to discretize ϕ into a finite set of values (say phi=30:150) and apply GA to the discretized problem. Then, take that solution and use it to initialize fseminf as applied to the original semi-infinite problem. Assuming GA does a decent job, you can reasonably expect the initial guess to be good enough so that the local optimization finds a global solution.

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

siddhesh rane 2017-10-22

I was thinking on same lines. Thanks.

请先登录，再进行评论。

Answer 2

John D'Errico 2017-10-22

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/362625-what-function-can-i-use-for-finding-global-optima-in-semi-infinite-constraint-optimization-problem#answer_287116

Why not? Only you know what is inside f, so being able to say, of course it can, this is impossible to know.

Of course, you seem to think that GA will truly give a global optimum. No numerical optimization tool can assure that on a completely blackbox function. All that can be done is to improve the probability that it will find a point at the global min.

The best thing that you can do is based on your own knowledge of the process, i.e., to provide good starting values. Any intelligence that you can add to the problem will be worth more than any magical computer algorithm.