There are some "visual" ways to think about I1 and I2 (the internal
evaluation criteria).

I1 is the "ball of string". Imagine that each context in a cluster is a
point. You want to connect all of the points to each other with pieces
of string that you will pull tight. The length of each string represents
the distance between the connected contexts, so you want that string to be
as small as possible (to have the highest possible similarity). So you
connect all the points to all of the other points with as little string as
possible. At the middle of the ball of string will like the centroid of
the cluster. It's not specifically used or represented in I1, but it is
there. Ideally you will want your ball of string to be round and small.

I2 is the "flower". Here the centroid is used, and it represents the
center of the flower. Each context is again a point, and the idea is to
connect each point to the centroid with a lovely yellow string, so that
you form a flower. The object is to minmize the length of each string
(whith represents distance) so you want to end up with a flower that is
nice and round.

So I1 represents a graph where the contexts are completely connected,
and I2 represents a graph where none of the contexts are directly
connected, they only connect to the centroid. So you can think of a ball
of string (for I1) and a flower (for I2). The object of the criterion
function is to find a cluster solution where each cluster is as small and
"tight" a ball of string or flower as as possible.

I find these sorts of visualization helpful to remember, quickly, the
differences in the criterion functions.

I hope it helps!
Ted

--
Ted Pedersen
http://www.d.umn.edu/~tpederse


-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf
_______________________________________________
senseclusters-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/senseclusters-users

Reply via email to