cs345-cl2

# cs345-cl2 - More Clustering CURE Algorithm Non-Euclidean...

This preview shows pages 1–13. Sign up to view the full content.

1 More Clustering CURE Algorithm Non-Euclidean Approaches

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
2 The CURE Algorithm ± Problem with BFR/k -means: ² Assumes clusters are normally distributed in each dimension. ² And axes are fixed --- ellipses at an angle are not OK. ± CURE: ² Assumes a Euclidean distance. ² Allows clusters to assume any shape.
3 Example: Stanford Faculty Salaries e e e e e e e e e e e h h h h h hh h h h h h h salary age

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
4 Starting CURE 1. Pick a random sample of points that fit in main memory. 2. Cluster these points hierarchically --- group nearest points/clusters. 3. For each cluster, pick a sample of points, as dispersed as possible. 4. From the sample, pick representatives by moving them (say) 20% toward the centroid of the cluster.
5 Example: Initial Clusters e e e e e e e e e e e h h h h h h hh h h h h h salary age

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
6 Example: Pick Dispersed Points e e e e e e e e e e e h h h h h h hh h h h h h salary Pick (say) 4 remote points for each cluster. age
7 Example: Pick Dispersed Points e e e e e e e e e e e h h h h h h hh h h h h h salary Move points (say) 20% toward the centroid. age

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
8 Finishing CURE ± Now, visit each point p in the data set. ± Place it in the “closest cluster.” ² Normal definition of “closest”: that cluster with the closest (to p ) among all the sample points of all the clusters.
9 Curse of Dimensionality ± One way to look at it: in large- dimension spaces, random vectors are perpendicular. Why? ± Argument #1: Lots of 2-dim subspaces. There must be one where the vectors’ projections are almost perpendicular. ± Argument #2: Expected value of cosine of angle is 0.

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
10 Cosine of Angle Between Random Vectors ± Assume vectors emanate from the origin (0,0,…,0). ± Components are random in range [-1,1]. ± (a 1 ,a 2 ,…,a n ).(b 1 ,b 2 ,…,b ) has expected value 0 and a standard deviation that grows as n. ± But lengths of both vectors grow as ± So dot product around n/ ( n * n) = 1/
11 Random Vectors --- Continued ± Thus, a typical pair of vectors has an angle whose cosine is on the order of 1/ n. ± As n -> , that’s 0; i.e., the angle is about 90°.

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
12 Interesting Consequence ± Suppose “random vectors are perpendicular,” even in non-Euclidean spaces.
This is the end of the preview. Sign up to access the rest of the document.

## This document was uploaded on 01/06/2012.

### Page1 / 42

cs345-cl2 - More Clustering CURE Algorithm Non-Euclidean...

This preview shows document pages 1 - 13. Sign up to view the full document.

View Full Document
Ask a homework question - tutors are online