2
$\begingroup$

Is there a solution to find the global optimum of a convex function , subject to non convex constraints?

For example:

$\min x_1^2 + x_2^2$

subject to

$x_2 + x_1^2 \geq 2$

$x_2 - x_1^2 \leq 3$

constraints

It's only an example; I am looking for a general method to find the optimal point in any convex function with non-convex constraints

  • 0
    What inequality constraints define the feasible region? – 2017-02-20

1 Answers 1

1

The two constraints here are actually both convex, which you can see by drawing out the shape and observing that circles (and lunes) are convex. Alternatively, you know that $x_1^2 + x_2^2$ is a convex function, so $(x_1 - 1)^2 + (x_2 - 1)^2$ is a convex function, and hence the inequality $(x_1 - 1)^2 + (x_2 - 1)^2 \leq 3$ is a convex region.

Also by drawing the regions out, you can see that the origin is included in the feasible set, and $(0, 0)$ clearly minimises $x_1^2 + x_2^2$, so the minimum is $0$.

  • 0
    I think i have chosen an incorrect example for my question, in general have can we solve a convext function subject to non-convex constraints? – 2017-02-20
  • 0
    @B.Mehta I disagree, the feasible region is not convex: If you draw a line segment between the bottom left "corner" and bottom right "corner", it passes through points that are not in the feasible set, therefore it is not convex. – 2017-02-20
  • 0
    @Kuifje The feasible region was originally the intersection of two circles, but has since been edited to a non-convex set, I agree. – 2017-02-20
  • 0
    @B.Mehta: all righty then ! – 2017-02-20
  • 0
    @B.Mehta I changed the constraints to show that, I am not looking for the current problem's solution. I Only want to know if there exist a solution for non-convex constraints . – 2017-02-21