Transcript Danny`s slides

The Canopies Algorithm
from
“Efficient Clustering
of High-Dimensional
Data Sets
with Application to Reference Matching”
Andrew McCallum, Kamal Nigam, Lyle H. Unger
Presented by Danny Wyatt
Record Linkage Methods
 As classification [Felligi & Sunter]



Data point is a pair of records
Each pair is classified as “match” or “not
match”
Post-process with transitive closure
 As clustering



Data point is an individual record
All records in a cluster are considered a
match
No transitive closure if no cluster overlap
Motivation
 Either way, n2 such evaluations must be
performed
 Evaluations can be expensive


Many features to compare
Costly metrics (e.g. string edit distance)
 Non-matches far outnumber matches
 Can we quickly eliminate obvious non-
matches to focus effort?
Canopies
 A fast comparison groups the data into
overlapping “canopies”
 The expensive comparison for full
clustering is only performed for pairs in
the same canopy
 No loss in accuracy if:
“For every traditional cluster, there exists
a canopy such that all elements of the
cluster are in the canopy”
Creating Canopies
 Define two thresholds


Tight: T1
Loose: T2
 Put all records into a set S
 While S is not empty




Remove any record r from S and create a
canopy centered at r
For each other record ri, compute cheap
distance d from r to ri
If d < T2, place ri in r’s canopy
If d < T1, remove ri from S
Creating Canopies
 Points can be in more than one canopy
 Points within the tight threshold will not
start a new canopy
 Final number of canopies depends on
threshold values and distance metric
 Experimental validation suggests that T1
and T2 should be equal
Canopies and GAC
 Greedy Agglomerative Clustering




Make fully connected graph with a node for
each data point
Edge weights are computed distances
Run Kruskal’s MST algorithm, stopping
when you have a forest of k trees
Each tree is a cluster
 With Canopies


Only create edges between points in the
same canopy
Run as before
EM Clustering
 Create k cluster prototypes c1…ck
 Until convergence



Compute distance from each record to each
prototype ( O(kn) )
Use that distance to compute probability of
each prototype given the data
Move the prototypes to maximize their
probabilities
Canopies and EM Clustering
 Method 1


Distances from prototype to data points only
computed within a canopies containing the
prototype
Note that prototypes can cross canopies
 Method 2
 Same as one, but also use all canopy
centers to account for outside data points
 Method 3

Same as 1, but dynamically create and
destroy prototypes using existing techniques
Complexity
 n : number of data points
 c : number of canopies
 f : average number of canopies covering a data
point
 Thus, expect fn/c data points per canopy
 Total distance comparisons needed becomes
2
  fn  2 


f
2
O c    O
n 
  c  
 c



Reference Matching Results
 Labeled subset of Cora data

1916 citations to 121 distinct papers
 Cheap metric


Based on shared words in citations
Inverted index makes finding that fast
 Expensive metric

Customized string edit distance between
extracted author, title, date, and venue
fields
 GAC for final clustering
Reference Matching Results
Method
F1
Error Precision
Recall
Minutes
Canopies
0.838 0.75%
0.735
0.976
7.65
Complete
GAC
0.835 0.76%
0.737
0.965
134.09
Author/Year
0.697 1.60%
0.559
0.926
0.03
none
1.99%
1.000
0.000
0.00
Discussion
 How do cheap and expensive distance
metrics interact?



Ensure the canopies property
Maximize number of canopies
Minimize overlap
 Probabilistic extraction, probabilistic
clustering

How do the two interact?
 Canopies and classification-based linkage

Only calculate pair data points for records in
the same canopy