Stability and Resampling Methods for Clustering

Cluster Stability Analysis Based on the Assessment of Individual Clusters

author: Patrice Bertrand, ENST Bretagne

You might be experiencing some problems with Your Video player.

*Slides*
0:00	Cluster stability analysis based on the assessment of individual clusters
0:26	Partition stability
3:02	Outline
4:34	Our notations
5:01	Stability based on sampling the data set
6:50	Artificial data set
7:13	Correlation similarity
7:49	Asymptotic results
8:32	Example of a unique minimizer
9:18	Example of instability from symmetry
9:29	Example (continuing)
9:41	Some issues
16:29	A proposal for measuring cluster stability w.r.t. cohesion and isolation
17:32	1. Stability measures
21:14	Isolation of a cluster
24:43	Isolation of a cluster 01
25:43	Isolation between two clusters
27:17	Other cluster features
29:12	Self learning the number of samples
30:35	p -value of a stability measure
33:28	”Optimal number” of clusters
34:14	2. Comparison with other validation measures
35:20	Artificial data set
35:37	Artificial data set 01
36:09	Stability measures and p-values
36:31	Stability measures and p-values (5-partition)
37:30	Iris data
38:09	Characterizing different types of unstability
38:54	3 symmetrical gaussians (continuing)
39:31	Stability measures
39:58	Data set #2: Uniform data set Partition: 3 clusters
40:19	Uniform data set (continuing)
41:19	Stability measures
41:23	Data set #3: 2 Gaussians with different variances Partition: 2 clusters
42:27	Stability measures
42:44	Data sets #4: 2 Gaussians with same variances Partition: 2 clusters
43:09	Data sets #4 (continuing)
43:50	Data sets #4 (continuing)
43:59	Data sets #5: 2 Gaussians with same variances Partition: 2 clusters
45:00	Data sets #5 (continuing)
47:04	Data set #6: Mixture of 1 Gaussian and 1 uniform law Partition: 3 clusters
47:28	Stability measures
47:48	Two individual scores
50:30	Membership scores of intermediary points
51:34	Filiation scores of intermediary points
52:37	Some conclusions and perspectives

Lecture rating

People found this lecture:
Worth seeing
because it is:
Valuable and informative
Well presented
Easily understandable
Acceptably recorded
You need to login to cast your vote.

Report a problem or upload files

If you have found a problem with this lecture or would like to send us extra material, articles, exercises, etc., please use our ticket system to describe your request and upload the data.
Enter your e-mail into the 'Cc' field, and we will keep you updated with your request's status.