Seperation of a Point and Convex Set

Click For Summary

Homework Help Overview

The discussion revolves around the separation of a point from a closed convex set, specifically addressing the properties of the perpendicular bisector and its relationship with the convex set. The problem involves demonstrating that no points of the convex set lie on the bisector or in the half-plane containing the external point.

Discussion Character

  • Exploratory, Conceptual clarification, Problem interpretation

Approaches and Questions Raised

  • Participants explore the implications of the convexity of set C and the geometric properties of the perpendicular bisector L. There are attempts to visualize the relationship between points on L and the closest point p in C. Questions arise about the nature of points far from the midpoint s on L and their relation to C.

Discussion Status

Some participants have offered geometric arguments regarding the properties of the convex set and the implications of the perpendicular bisector. There is an ongoing exploration of the relationship between closed and open half-planes in a related problem, with differing opinions on the necessity of their definitions.

Contextual Notes

Participants are considering the definitions of closed and open half-planes in the context of convex sets and their intersections, questioning the implications of these definitions on the overall problem. There is a focus on ensuring that the intersection does not include points outside of C.

e(ho0n3
Messages
1,349
Reaction score
0
[SOLVED] separation of a Point and Convex Set

Homework Statement


Let C be a closed convex set and let r be a point not in C. It is a fact that there is a point p in C with |r - p| l<= |r - q| for all q in C.

Let L be the perpendicular bisector of the line segment from r to p. Show that no point of C lies on L or in the half-plane, determined by L, which contains r.

2. The attempt at a solution
The point p happens to be the closest point in C to r. I can imagine how the point s in the middle of the line segment from r to p and some of the points near s on L could not be in C because then there would be a closer point to r in C than p. But what about points far away from s? I'm guessing the fact that C is closed and convex comes into play at this point, but how?
 
Physics news on Phys.org
I believe the idea you want to pursue is this:

Suppose there is a point x on L that is in C and consider the line L' joining x and p. Then all points of that line are in C because C is convex, and the orthogonal projection of r on that line is not p because if it were, the line L' would be parallel to L and hence never meet it.

(the orthogonal projection of r on L' is the point of L' that minimizes the distance btw L' and r. since L' is in C, we have a contradiction is that o.p. is not p)
 
Last edited:
That is a neat argument. If I imagine L', I can visualize how a point on it very close to p will be closer to r than p. This has now been reduced to a geometry problem which I think I can solve. Thanks a lot.
 
I have another problem: Show that each closed convex set is the intersection of all the closed half-planes that contain it.

For a point r not in C, I can always determine a line L as described in the previous posts which consequently determines an open half-plane that contains C but not r. If I do this for every point r not in C and take the intersection of these open half-planes, the intersection will not contain any of the points not in C. In other words, the intersection will just contain C.

The only problem here is that the half-planes are open. Perhaps the author meant to write open instead of closed. What do you think?
 
I think an arbitrary intersection of half spaces will be closed, so no prob.

just like the intersection of all (1/n,infty) is [0,,infty).
 
I guess it doesn't really matter whether the half-planes containing C are open or closed because either way, they will exclude their corresponding r's and so the intersection will always end up having only points in C.
 
just an edit to my post: I didn't meant an arbitrary intersection of half space, but rather the intersection of those half-space.
 

Similar threads

Replies
4
Views
2K
  • · Replies 8 ·
Replies
8
Views
2K
  • · Replies 9 ·
Replies
9
Views
3K
  • · Replies 8 ·
Replies
8
Views
3K
  • · Replies 1 ·
Replies
1
Views
4K
  • · Replies 1 ·
Replies
1
Views
7K
  • · Replies 9 ·
Replies
9
Views
4K
  • · Replies 3 ·
Replies
3
Views
2K
  • · Replies 6 ·
Replies
6
Views
1K
  • · Replies 4 ·
Replies
4
Views
3K