1. Clustering is one of the most widely used tools for exploratory data analysis. Social Sciences...
-
Upload
arnold-collins -
Category
Documents
-
view
213 -
download
1
Transcript of 1. Clustering is one of the most widely used tools for exploratory data analysis. Social Sciences...
Towards Theoretical Foundations of Clustering
Margareta Ackerman
University of Waterloo
1
Clustering is one of the most widely used tools
for exploratory data analysis. Social Sciences Biology Astronomy Computer Science ….
All apply clustering to gain a first understanding of the structure of large data sets.
The Theory-Practice Gap
2
The Theory-Practice Gap
“While the interest in and application of cluster analysis has been rising rapidly, the abstract nature of the tool is still poorly understood” (Wright, 1973)
“There has been relatively little work aimed at reasoning about clustering independently of any particular algorithm, objective function, or generative data model” (Kleinberg, 2002)
Both statements still apply today. 3
Clustering aims to assign data into groups of similar items
Beyond that, there is very little consensus on the definition of clustering
Inherent Obstacles:Clustering is ill-defined
4
• Clustering is inherently ambiguous– There may be multiple reasonable
clusterings– There is usually no ground truth
• There are many clustering algorithms with different (often implicit) objective functions
Inherent Obstacles
5
• Previous work• Clustering algorithm selection• Characterization of Linkage-Based clustering
– Sketch of proof– Hierarchical algorithms that are not linkage-
based • Conclusions and future work
6
Outline
• Clustering in the weighted setting (Wright, ‘73)
• Axioms of clustering distance functions (Meila, ACM ‘05)
• Impossibility result (Kleinberg, NIPS ‘02)
• Rebuttal to impossibility result (Ackerman & Ben-David, NIPS ‘08)
7
Previous Work Towards a General Theory: Axiomatizing clustering
• Conditions for efficiently uncovering the target clustering [(Balcan, Blum, and Vempala, STOC ‘08),(Balcan, Blum and Gupta, SODA ‘09)]
• Theoretical study of clusterability (Ackerman & Ben-David, AISTATS ‘09)]. Notions of clusterability are pairwise distinct Data sets that are more clusterable are
computationally easier to cluster well.
8
Previous Work Towards a General Theory: Clusterability
• Previous work• Clustering algorithm selection• Characterization of Linkage-Based clustering
– Sketch of proof– Heirarchical algorithms that are not linkage-
based• Conclusions and future work
9
Outline
There are a wide variety of clustering algorithms, which often produce very different clusterings.
Clustering Algorithm Selection
10
How should a user decide which algorithm to use for
a given application?
Users rely on cost related considerations: running
times, space usage, software purchasing costs, etc…
There is inadequate emphasis on
input-output behaviour
Clustering Algorithm Selection
11
12
Radical Differences in Input/Output Behavior of Clustering Algorithms
13
Radical Differences in Input/Output Behavior of Clustering Algorithms
We propose a framework that lets a user utilize prior knowledge to select an algorithm
• Identify properties that distinguish between different input-output behaviour of clustering paradigms
• The properties should be:1) Intuitive and “user-friendly”2) Useful for distinguishing clustering
algorithms
Our Framework for Clustering Algorithm Selection
14
• The long-term goal is to construct a large property-based classification for many useful clustering algorithms
• This would facilitates the application of prior knowledge.
• Enables users to identify a suitable algorithm without the overhead of executing many algorithms
• This framework helps understand behaviour of existing and new algorithms
Our Framework for Clustering Algorithm Selection
15
Taxonomy of Partitional Algorithms(Ackerman, Ben-David, Loker, NIPS 2010)
Local OuterCon.
InnerCon.
Refinm.Preserv
OrderInv.
OuterRich.
ScaleInv.
Iso.Inv.
Single linkage
Average linkage
Complete linkage
K-means K-median Min-Sum Ratio-cut Normalized-cut
16
Axioms VS Properties
Local OuterCon.
InnerCon.
Refinm.Preserv
OrderInv.
OuterRich.
ScaleInv.
Iso.Inv.
Single linkage
Average linkage
Complete linkage
K-means K-median Min-Sum Ratio-cut Normalized-cut
Properties Axioms
17
Characterization of Linkage-Based Clustering(Ackerman, Ben-David, Loker, COLT 2010)
Local OuterCon.
InnerCon.
Refinm.Preserv
OrderInv.
OuterRich.
ScaleInv.
Iso.Inv.
Single linkage
Average linkage
Complete linkage
K-means K-median Min-Sum Ratio-cut Normalized-cut
18
Characterization of Linkage-Based Clustering(Ackerman, Ben-David, Loker, COLT 2010)
The 2010 characterization applies in the partitional setting, by using the k-stopping criteria.
This characterization distinguished linkage-based algorithms from other partitional algorithms.
Local OuterCon.
InnerCon.
Refinm.Preserv
OrderInv.
OuterRich.
ScaleInv.
Iso.Inv.
Single linkage
Average linkage
Complete linkage
19
• Propose two intuitive properties that uniquely indentify hierarchical linkage-based clustering algorithms.
• Show that common hierarchical algorithms, including bisecting k-means, cannot be simulated by any linkage-based algorithm
Characterizing Linkage-Based Clustering in the Heirarchical Setting
(Ackerman and Ben-David, IJCAI 2011)
20
• Previous work• Clustering algorithm selection• Characterization of Linkage-Based clustering
– Sketch of proof– Hierarchical algorithms that are not linkage-
based• Conclusions and future work
21
Outline
C_i is a cluster in a dendrogram D if there exists a node in the dendrogram so that C_i is the set of its leaf descendents.
Formal Setup: Dendrograms and clusterings
22
C = {C1, … , Ck} is a clustering in a dendrogram D if
– Ci is a cluster in D for all 1≤ i ≤ k, and
– Clusters are disjoint
Formal Setup: Dendrograms and clusterings
23
Formal Setup: Heirarchical clustering algorithm
A Hierarchical Clustering Algorithm A maps
Input: A data set X with a dissimilarity function d, denoted (X,d)
toOutput: A dendrogram of X
24
• Create a leaf node for every elements of X
Linkage-Based Algorithm
Insert image
25
• Create a leaf node for every elements of X
• Repeat the following until a single tree remains:– Consider clusters represented by the remaining root nodes.
Linkage-Based Algorithm
26
• Create a leaf node for every elements of X
• Repeat the following until a single tree remains:– Consider clusters represented by the remaining root nodes.
Merge the closest pair of clusters by assigning them a common parent node.
Linkage-Based Algorithm
27
?
• The choice of linkage function distinguishes between different linkage-based algorithms.
• Examples of common linkage-functions– Single-linkage: shortest between-cluster distance– Average-linkage: average between-cluster distance– Complete-linkage: maximum between-cluster distance
Examples of Linkage-Based Algorithms
X1 X2
28
Locality Informal Definition
If we select a set of disjoint clusters from a dendrogram, and run the algorithm on the union of these clusters, we obtain a result that is consistent with the original dendrogram.
D = A(X,d) D’ = A(X’,d)X’={x1, …, x6}
29
Locality Informal Definition
If we select a set of disjoint clusters from a dendrogram, and run the algorithm on the union of these clusters, we obtain a result that is consistent with the original dendrogram.
D = A(X,d) D’ = A(X’,d)X’={x1, …, x6}
30
A(X,d)
C
C on dataset (X,d)C on dataset (X,d’)
Outer-consistent change
31
Outer Consistency
If A is outer-consistent, then A(X,d’) will also include the clustering C.
Theorem (Ackerman & Ben-David, IJCAI 2011):
A hierarchical clustering algorithm is
Linkage-Based if and only if
it is Local and Outer-Consistent.
Characterization of Linkage-Based Clustering
32
• Previous work• Clustering algorithm selection• Characterization of Linkage-Based clustering
– Sketch of proof– Heirarchical algorithms that are not linkage-
based• Conclusions and future work
33
Outline
Every Linkage-Based hierarchical clustering algorithm is Local and Outer-Consistent.
The proof is quite straightforward.
Easy Direction of Proof
34
If A is Local and Outer-Consistent, then A is Linkage-Based.
To prove this direction we first need to formalize Linkage-Based clustering, by formally defining what is a Linkage Function.
Interesting Direction of Proof
35
A Linkage Function is a function
l:{(X1, X2 ,d): d is a distance function over X1uX2 }→ R+
that satisfies the following:
What Do We Expect From Linkage Functions?
- Representation independence: Doesn’t change if we re-label data - Monotonicity: if we increase edges that go between X1 and X2,
then l(X1, X2 ,d) doesn’t decrease.
(X1uX2,d)
X1 X2
36
Recall direction: If A satisfies Outer-Consistency and Locality, then A is Linkage-Based.
Goal: Define a linkage function l so that the linkage-based clustering based on l outputs A(X,d) (for every X and d).
Sketch of proof
37
• Define an operator <A :
(X,Y,d1) <A (Z,W,d2) if when we run A on (XuYuZuW,d), where d extends d1 and d2, X and Y are merged before Z and W.
Sketch of proof
A(X,d)
Z W X Y
• Prove that <A can be extended to a partial ordering
• Use the ordering to define l
38
Sketch of proof continue:Show that <A is a partial ordering
We show that <A is cycle-free.
Lemma: Given a hierarchical algorithm A that is Local and Outer-Consistent, there exists no finite sequence so that
(X1,Y1,d1) <A …. <A(Xn,Yn,dn) <A (X1,Y1,d1).
39
• By the above Lemma, the transitive closure of <A is a partial ordering.
• This implies that there exists an order preserving function l that maps pairs of data sets to R+.
• It can be shown that l satisfies the properties of a Linkage Function.
Sketch of proof (continued…)
40
• Previous work• Clustering algorithm selection• Characterization of Linkage-Based clustering
– Sketch of proof– Hierarchical algorithms that are not linkage-
based• Conclusions and future work
41
Outline
Hierarchical but Not Linkage-Based
P -Divisive algorithms construct dendrograms top-downusing a partitional 2-clustering algorithm P to split nodes.
42
Apply partitional clustering P Ex. k-means for k=2
Hierarchical but Not Linkage-Based
A partitional 2-clustering algorithm P is Context Sensitive if there exist d d’⊂ so that
P({x,y,z},d) = {x, {y,z}} and P({x,y,z,w} ,d’)= {{x,y}, {z,w}}.
Ex. K-means, min-sum, min-diameter.
43
Theorem [Ackerman & Ben-David, IJCAI ’11]:
If P is context-sensitive, then the P –divisive algorithm fails the locality
property.
Hierarchical but Not Linkage-Based
• The input-output behaviour of some natural divisive algorithms is distinct from that of all linkage-based algorithms.
• The bisecting k-means algorithm, and other natural divisive algorithms, cannot be simulated by any linkage-based algorithm.
44
Conclusions
• We present a new framework for clustering algorithm selection
• Provide a property-based classification of common clustering algorithms
• Characterize linkage-based clustering in terms of two natural properties
• Show that no linkage-based algorithm can simulate some natural divisive algorithms
45
What’s Next?• Our approach to selecting clustering algorithms can
be applied to any clustering application (ex. phylogeny).
• Classify applications in terms of their clustering needs– Target research on common clustering needs or specific
applications– Identify when results are relevant to specific applications
• Bridging the gap in other clustering settings (ex. clustering with a “noise cluster”)
• Axioms of clustering algorithms46