1

How can we formulate a linear programme that tells us whether an arbitrary point x[ j ] ∈ X, where X = {x1, ... ,xn} ⊂ Rn is an extreme point of the convex hull of X, that is conv(X)?

According to the solution of this linear programme, we should be able to claim that 'yes, x[ j ] is an extreme point', or 'no'.

Well, what I've in my mind is something like that:

{min: 0} s.t. x[ j ] = Σi ( a[ i ] * x[ i ] ); i ∈ {1, ... ,k}, ∀ j ∈ {1, ... ,k}

If such a[ i ] s exist, that means x[ j ] is a linear combination of other x's, which seems like a violation to the definition of an extreme point.

However, I believe this LP does not cover the whole context. i.e., what if we select an x[ j ] which is located inside the conv(X) (not on the edges) and is not a linear combination of others. Then the model will result in a fallacious outcome. It seems to me that, above model would be fine iff the selected x [ j ] stands on the edges of conv(X).

Thanks.

Izzy
  • 21
  • 4
  • You should include what you have tried. There plenty of examples on the web for Convex Hull routines. This looks like a copy paste of a homework assign with no thought. – Matt Nov 14 '16 at 17:30
  • @Matt, please see the edit on what I've tried. I've been researching and thinking about it for hours. Asking here, was my last effort, not the first. – Izzy Nov 14 '16 at 18:13

1 Answers1

1

You are quite close. A points belongs to the convex hull of a finite set of points if and only if it can be written as a convex combination of these points. Also, a point is an extreme point only if there are no two other points for which it can be written as a convex combination of them.

Let the point of interest be x_k. Then, the following linear program will do the job:

enter image description here

where x_{ik} is the i-th coordinate of the point you want to check (point k). Note that this point should be one of the points we include in the right-hand side of the equation (i.e., the problem will always have at least one solution were lambda_k = 1 and all other lambdas will equal 0.

If that point is an extreme point, then the only solution you will get is lambda_k=1, other lambdas = 0. Otherwise, a different solution (with smaller lambda_k) will pop up.

(Note that in your problem description both the number of points and the dimensions are n), therefore the corresponding indexes (j and i) run from 1 to n.

I hope this helps!

Community
  • 1
  • 1
Ioannis
  • 5,238
  • 2
  • 19
  • 31