Subspace Clustering Ensemble Clustering Subspace Clustering, - - PowerPoint PPT Presentation

LUDWIG- MAXIMILIANS- UNIVERSITÄT MÜNCHEN DATABASE SYSTEMS GROUP INSTITUTE FOR INFORMATICS MÜNCHEN GROUP INFORMATICS

Subspace Clustering Ensemble Clustering Subspace Clustering, Ensemble Clustering, Alternative Clustering, Multiview Clustering: What Can We Learn From Each Other?

MultiClust@KDD 2010

Hans-Peter Kriegel, Arthur Zimek

Ludwig-Maximilians-Universität München Munich, Germany http://www.dbs.ifi.lmu.de {kriegel, zimek}@dbs.ifi.lmu.de

DATABASE SYSTEMS GROUP

Outline

GROUP

• 1. Subspace Clustering
• 2. Ensemble Clustering
• 3. Alternative Clustering
• 4. Multiview Clustering
• 5. Discussion

Kriegel/Zimek: What can we learn from each other? (MultiClust@KDD 2010)

2

DATABASE SYSTEMS GROUP

Subspace Clustering

GROUP

• Task: identify clusters of similar objects

i il it d fi d t t i b f th d t

• similarity defined w.r.t. a certain subspace of the data space
• different subspaces for different clusters

Kriegel/Zimek: What can we learn from each other? (MultiClust@KDD 2010)

3

DATABASE SYSTEMS GROUP

Subspace Clustering

GROUP

• Subspaces: different

– selection – weighting combination – combination

• f attributes
• learn subspace and clustering
• learn subspace and clustering

simultaneously (interdepency)

• strategies:

strategies:

– top-down (learn spatial characteristics

• f initially built sets of objects)

evant attribute

– bottom-up (learn 1-d clusters, combine them to 2-d clusters, etc. (APRIORI)) => many irrelevant clusters

irrele

Kriegel/Zimek: What can we learn from each other? (MultiClust@KDD 2010)

a y e e a c us e s

4

relevant attribute/ relevant subspace

DATABASE SYSTEMS GROUP

Ensemble Clustering

GROUP

• basic idea: combine different clusterings to obtain one

i l li bl l t i single, more reliable clustering

• tasks:

h t t di l t i – how to create diverse clusterings – how to combine different clusterings

• induce diversity of clusterings
• induce diversity of clusterings

– use different feature-subsets – use different database subsets – use different clustering algorithms

• correspondence between clusterings

– useful for judging on redundancy of clusters? – a lot of different answers – but: could it not be that different clusterings are just different yet both meaningful?

Kriegel/Zimek: What can we learn from each other? (MultiClust@KDD 2010)

clusterings are just different, yet both meaningful?

5

DATABASE SYSTEMS GROUP

Alternative Clustering

GROUP

• given a clustering, use diversity or non-redundancy as a

t i t t fi d diff t l t i constraint to find a different clustering

• techniques:

bl t h i – ensemble techniques – use different subspaces

• relationship to subspace clustering:
• relationship to subspace clustering:

– subspace clustering can learn from the treatment of non-redundancy – alternative clustering can learn to allow for a certain level of g redundancy

Kriegel/Zimek: What can we learn from each other? (MultiClust@KDD 2010)

6

DATABASE SYSTEMS GROUP

Multiview Clustering

GROUP

• seek different clusterings in different subspaces
• special case of alternative clustering?

– constraint: orthogonality of subspaces

• special case of subspace clustering?

– allowing maximal overlap of clusters seeking minimally redundant clusters by accommodating different – seeking minimally redundant clusters by accommodating different concepts

• emphasizes the observation known from subspace

p p clustering: highly overlapping clusters in different subspaces need not g y pp g p be redundant nor meaningless

Kriegel/Zimek: What can we learn from each other? (MultiClust@KDD 2010)

7

DATABASE SYSTEMS GROUP

Discussion

GROUP

subspace clustering

l diff t l t i diff t

ensemble clustering

l diff t b h ll

• goal: different clusters in different

subspaces

• problem: redundancy of clusters

(same clusters reported for

• goal: different subspaces shall

induce the same clusters

• problem: correspondence of

clusterings? What about actually (same clusters reported for different subspaces) clusterings? What about actually different clusterings?

? multiview clustering

• goal: find different cluster

alternative clustering

• goal: given a clustering find a
• goal: find different cluster

concepts in different subspaces

• problem: balance between

admissible overlap of clusters and

• goal: given a clustering, find a

different clustering

• problem: which level of

redundancy is admissible?

Kriegel/Zimek: What can we learn from each other? (MultiClust@KDD 2010)

8

admissible overlap of clusters and difference between concepts redundancy is admissible?

DATABASE SYSTEMS GROUP

Discussion

GROUP

• how should we treat diversity of clustering solutions?

– should diverse clusterings always be unified (ensemble)? – should diverse clusterings always be unified (ensemble)? – under which conditions is a unification of diverse clusterings meaningful?

• can we learn from diversity itself?

– again ensemble: exceptional clustering in one subspace will be t b d d l t ld it t b i ll i t ti ?

• utnumbered and lost – could it not be especially interesting?
• how to treat redundancy (esp. overlap)?

when does a cluster qualify as redundant w r t another cluster when – when does a cluster qualify as redundant w.r.t. another cluster, when does it represent a different concept (despite a certain overlap)?

hi h d d subspace clustering alternative clustering

?

l d d

• how to assess similarity between clustering solutions?

– possible overlap between clusters makes this problem really difficult

high redundancy low redundancy

Kriegel/Zimek: What can we learn from each other? (MultiClust@KDD 2010)

poss b e o e ap be ee c us e s a es s p ob e ea y d cu – no simple mapping

9