Convex Hull Problem from Monthly
From: Bill97 (nospam)
Date: 12/16/04
- Next message: Stephen Harris: "Re: Poll: Are PCs Turing Machines?"
- Previous message: EDWARD HYMAN: "A Laplace Transform Integral"
- Next in thread: Stephen J. Herschkorn: "Re: Convex Hull Problem from Monthly"
- Reply: Stephen J. Herschkorn: "Re: Convex Hull Problem from Monthly"
- Reply: ArtflDodgr: "Re: Convex Hull Problem from Monthly"
- Reply: Jim Ferry: "Re: Convex Hull Problem from Monthly"
- Messages sorted by: [ date ] [ thread ]
Date: Wed, 15 Dec 2004 20:03:36 -0500
This concerns problem #11120 given in the current (December, 2004) copy of
the American Mathematical Monthly, on page 915.
It is asked to show that the expected number of vertices of the convex hull
of n points sampled from the standard normal distribution in the plane is
2*sqrt(2*pi*ln n). (The probability density function for the normal
distribution is 1/2pi * e^(-(x^2+y^2)/2).
A couple of my colleagues were discussiong this problem today. Let us just
consider the case n=3 right now:
It seems intuitive to me that the probability of 3 sampled points being
collinear is 0, so that the expected number of vertices when n=3 would be 3.
Moreover, the formula yields the unlikely value 2*sqrt(2*pi*ln 3) = 5.25!
Perhaps someone would be willing to help me to clarify my understanding of
the problem.
Thanks,
Bill
- Next message: Stephen Harris: "Re: Poll: Are PCs Turing Machines?"
- Previous message: EDWARD HYMAN: "A Laplace Transform Integral"
- Next in thread: Stephen J. Herschkorn: "Re: Convex Hull Problem from Monthly"
- Reply: Stephen J. Herschkorn: "Re: Convex Hull Problem from Monthly"
- Reply: ArtflDodgr: "Re: Convex Hull Problem from Monthly"
- Reply: Jim Ferry: "Re: Convex Hull Problem from Monthly"
- Messages sorted by: [ date ] [ thread ]