quasi
Posts:
9,078
Registered:
7/15/05
|
|
Re: Question: Centroid given a distance metric
Posted:
Feb 12, 2013 3:49 AM
|
|
Andrey Savov wrote
>
>In other words, all I can do is, given 2 points, tell the
>distance between them. I'm looking for known algorithms to
>locate a point that minimizes all distances between itself
>and the original points.
Minimizes all distances? That makes no sense.
What you previously said and what I think you meant to say
above is that you want to find a point which minimizes the
sum of the squares of the distances from the point to the
points of the given set.
For a general distance function, the sum of squares of the
distances may have no geometric significance whatsoever, so
calling a point which minimizes that sum "a centroid" is, in
my opinion, a poor choice of terminology.
>Pretty sure I can prove existence and uniqueness of these
>assuming only metric definition:
>
<http://en.wikipedia.org/wiki/Metric_(mathematics)#Definition>
Can you prove that a metric, regarded as a function from
R^n x R^n -> R^n, is continuous? If not, existence of the point
you seek may not be guaranteed.
And even if you assume the metric is continuous, uniqueness
may fail. For example, with the discrete metric where
d(p,q) = 1 for all distinct points p,q, any point of the
original set qualifies as a point which minimizes the sum of
the squares of the distances to the set.
quasi
|
|
