Re: plot cluster of points without knowing the coordinates.



Paul K. wrote:
Hi!
Is it possible to plot a cluster of points in matlab in 2D given that
we have complete access to all the pairwise distances between points
in a cluster?

For example we have the set of points {p1,p2,p3,p4} with distances:

p1p2=3
p1p3=5
p2p3=1

Say we choose a convention where p1 is the origin of a coord system, and p2 is on the X axis so we have a reference frame. This can later be rotated and/or translated as desired. We now have p1=(0,0) and p2=(3,0) for the example data.

The distance p1-p3=5 defines a circle centered at p1=(0,0) of radius 5. The distance p2-p3=1 defines a circle centered at p2=(3,0) of radius 1. These circles do not intersect, so the example data is inconsistent and represents an impossible situation.

Say the data had p2-p3=2. Then the p1-p3 circle and the p2-p3 circle are tangent at p3=(5,0). That solution would be unique, with p1, p2, and p3 collinear.

Now say the data had 8 > p2-p3 > 2. Then the circles would intersect at two points. Which point is "really" p3?

Do additional points help resolve ambiguity? If we introduce distances p1-p4, p2-p4, and p3-p4, we *could* have three circles which, for consistent data representing a real situation, resolves to a single point for p4 *if* p1, p2, and p3 are not collinear. But we don't have only three circles. There were two possible circles for p3-p4.

OK, then have a convention that the first ambiguity is arbitrarily resolved (effectively choosing one of the two mirror solutions). Then it appears a solution is possible by triple-wise circle intersections, although it would be necessary to handle special cases (inconsistency, tangency, collinearity) for all triples, and probably have to deal with numerical tolerance issues...

....Ron Capelli (Austin)
.



Relevant Pages

  • Re: Average Distance to Circumference
    ... placed on a random point in the circle and immediately begins crawling ... different distances are "cancelled out" by the different distances on ... insect takes off at an angle theta to the x-axis (where theta = 0 ... If you don't know integral calculus, ...
    (sci.math)
  • Re: Average Distance to Circumference
    ... assume we have a circle with a radius of ten meters. ... different distances are "cancelled out" by the different distances on ... is to the right, and theta increases counterclockwise). ... If you don't know integral calculus, ...
    (sci.math)
  • Re: Volume of a section of a sphere
    ... A unit sphere is cut into 4 sections by two perpendicular planes at ... In two dimensions you are looking at the unit circle with two perpendicular ... lines cutting it at distances a and b. ...
    (sci.math)
  • Re: Great Circle Antenna Pointing Question
    ... It is an Azimuthal-Equidistance Map Generator v.3.7 ... >A bit confused over Great Circle directions. ... >Lots that calculate distances, but couldn't locate any that show distances, ...
    (rec.radio.shortwave)
  • Re: How far can you cluster (SCS) unsupported over a WAN?
    ... > protocols not just IP) a simulated LAN over IP. ... At long distances, the round-trip time is dominated by latency due to ... the OpenVMS Cluster Software SPD ... or 500 miles if you purchase the Disaster ...
    (comp.os.vms)