Multilinear Algebra for Analyzing Data with Multiple Linkages Tamara G. Kolda plus: Brett Bader,...
-
Upload
elle-buttler -
Category
Documents
-
view
217 -
download
0
Transcript of Multilinear Algebra for Analyzing Data with Multiple Linkages Tamara G. Kolda plus: Brett Bader,...
Multilinear Algebra for Analyzing Data with Multiple Linkages
Tamara G. Kolda
plus: Brett Bader, Danny Dunlavy, Philip Kegelmeyer
Sandia National Labs
TRICAP 2006, Chania, Greece, June 4-9, 2006
Tamara G. Kolda – TRICAP – June 6, 2006 - p.2
Linear Algebra for Data with Linkages
Circle-Square Matrix
Circle-Circle Co-Link Matrix Square-Square Co-Link Matrix
SVD Rank-k Approximation (k=2)
Tamara G. Kolda – TRICAP – June 6, 2006 - p.3
Latent Semantic Indexing (LSI)
for Text Retrieval
• S. T. Dumais, G. W. Furnas, T. K. Landauer, S. Deerwester, and R. Harshman. Using latent semantic analysis to improve access to textual information. In CHI '88, pp. 281–285, 1988
• S. C. Deerwester, S. T. Dumais, T. K. Landauer, G. W. Furnas, and R. A. Harshman. Indexing by latent semantic analysis. J. Am. Soc. Inform. Sci., 41(6):391–407, 1990
• M. W. Berry, S. T. Dumais, and G. W. O'Brien. Using linear algebra for intelligent information retrieval. SIAM Rev., 37(4):573–595, 1995
Term-Document Matrix
“Car Service”Query
SMART Retrieval SystemG. Salton (1971)
LSI S. Dumais et al. (1988)
Terms
Documents
car
repair
service
military
d1
d2
d3
Tamara G. Kolda – TRICAP – June 6, 2006 - p.4
Applications of LSI
Terms
Documents
car
repair
service
military
d1
d2
d3
Graph the Results using U2 and V2 Term-Document Similarities
car
service
military
repair
d1 d2 d3
car
service
military
repaircar service military repair
Term-Term
Document-Document
Tamara G. Kolda – TRICAP – June 6, 2006 - p.5
Caveats for LSI
• How to use • Term-document matrix weighting is critical!
Local WeightLog
fij = frequency
Global Term WeightInverse Document Frequency
N = total docsni = # docs with term i
Normalization Factor“Cosine”
Tamara G. Kolda – TRICAP – June 6, 2006 - p.6
Citation/Link Analysis(Same Nodes)
1
2
3
4
Link Matrix
Co-Citation Matrix
Co-Reference Matrix
Hub Scores
Authority Scores
Doc 3 is the most important authority!
Doc 1 is the most important hub!
J. M. Kleinberg. Authoritative sources in a hyperlinked environment. J. ACM, 46(5):604–632, 1999.
Examples:Citation data,
Web links
Tamara G. Kolda – TRICAP – June 6, 2006 - p.7
Multiple Links?
Suppose the connections between nodes are “labeled”
in some fashion.
In other words, we have meta-data on the
connections.
Can we somehow use
multilinear algebra for link
analysis?
Tamara G. Kolda – TRICAP – June 6, 2006 - p.8
PARAFAC
• PARAFAC = Parallel Factors
• aka. CANDECOMP = Canonical Decomposition
• Higher-order analogue of the SVD
• Columns of A, B, and C are not orthonormal
• If R is minimal, then R is called the rank of the tensor (Kruskal 1977)
• Can have rank(X) > min{I,J,K}
• Often guaranteed to be a unique rank decomposition!
=A
I x R
B
J x RI x J x K
C
K x R
R x R x R
= + + + …I
•R. A. Harshman. Foundations of the PARAFAC procedure: models and conditions for an “explanatory” multi-modal factor analysis. UCLA working papers in phonetics, 16:1–84, 1970•J. D. Carroll and J. J. Chang. Analysis of individual differences in multidimensional scaling via an N-way generalization of `Eckart-Young' decomposition. Psychometrika, 35:283–319, 1970.
Tamara G. Kolda – TRICAP – June 6, 2006 - p.9
Many ways to write PARAFAC
J. B. Kruskal. Three-way arrays: rank and uniqueness of trilinear decompositions, with application to arithmetic complexity and statistics. Linear Algebra Appl., 18(2):95–138, 1977.
“Kruskal Operator”
Easy to write N-way case:
“Tucker Operator”
Tamara G. Kolda – TRICAP – June 6, 2006 - p.10
Properties of the Kruskal Operator
PARAFAC core for a Tucker decomposition:
Matricize (arbitrary map of indices to rows and columns):
Mode-n matricize:
Norm of a PARAFAC decomposition:
Tamara G. Kolda – TRICAP – June 6, 2006 - p.11
PARAFAC for sparse data &
approximations• Our interest in the mathematical operations is motivated on two fronts
(1) Sparse computations
(2) Using tensor decompositions for approximation
• Ex: Considering how to efficiently implement PARAFAC-ALS for sparse data
• Can PARAFAC be used for the best rank-k approximation, rather than finding an exact decomposition (excepting noise)
What does it even mean in this case??
Tamara G. Kolda – TRICAP – June 6, 2006 - p.12
Multilink Analysis using PARAFAC
• Quick Review: Tensors for Web Link Analysis page x page x anchor text (TOPHITS)
• New work: Tensors for Publication Data Analysis Case 1: doc x doc x similarity
Case 2: term x doc x author (HO-LSA??)
Tamara G. Kolda – TRICAP – June 6, 2006 - p.13
TOPHITS: PARAFAC for Web Link Analysis
A set of four hyperlinked web pages
Graph representation shows basic connectivity
Labeled edges capture context
Tamara G. Kolda – TRICAP – June 6, 2006 - p.14
Analyzing Publication Data:Doc x Doc x Similarity
Representation
Tamara G. Kolda – TRICAP – June 6, 2006 - p.15
Computing Different Doc-Doc Similarities
Computing term-based similarities (k=1,2,3)
Computing author similarities (k=4)
Enforces sparseness!
• 5022 papers• 16617 unique
terms (ignoring stop words, words with length less than 3 or greater than 30 characters, and words that appear less than 2 times)
Titles: 5164 Abstracts: 15752 Keywords: 5248
• 6891 authors
• 2659 citations
Tamara G. Kolda – TRICAP – June 6, 2006 - p.16
PARAFAC for Doc x Doc x Similarity
• H = “hubs”
• A = “authorities”
• C = “connections”
• Rank-30 decomposition
Central idea:Each triplet provides a core “grouping” of the
data, i.e., a specific topic.
Tamara G. Kolda – TRICAP – June 6, 2006 - p.17
Sample: Grouping 1
Tamara G. Kolda – TRICAP – June 6, 2006 - p.18
Sample: Grouping 10
Tamara G. Kolda – TRICAP – June 6, 2006 - p.19
Applications of the [H,A,C]
Decomposition• Latent document similarities
Calculate S = ½ HHT + ½ AAT
• Analyzing a body of work ch = hub centroid, ca = authority centroid
s = ½ H ch + ½ A ca
• Disambiguation (EXAMPLE) Calculate centroids using A (could also use H or A+H)
Calculate simiarlities of centroids
• Journal predicition Use matrix A as features for input to a decision tree
ensemeble classifier
Tamara G. Kolda – TRICAP – June 6, 2006 - p.20
Example of Disambiguation
ResultsTwo authors with missing middle
initials.
3 possible matches
Matrix of Similarities
Tamara G. Kolda – TRICAP – June 6, 2006 - p.21
Analyzing Publication Data:Term x Doc x Author
Representation
term
doc auth
or
Form tensor X as:
Element (i,j,k) is nonzero only if author k wrote document j using term i.
767 documents2251 terms
1072 authors59738 nonzeros
Terms must appear in at least 3 documents and no
more than 10% of all documents. Moreover, it
must have at least 2 characters and no more
than 30.
Tamara G. Kolda – TRICAP – June 6, 2006 - p.22
Different Graph Interpretations for
Term x Doc x Author
• term-doc with author links
• term-author with doc links
• author-doc with term links
• term-doc-author with links
Term
DocDifferent author links represented
by different colors
Tamara G. Kolda – TRICAP – June 6, 2006 - p.23
Author Data is Too Sparse
term
doc
auth
or
Result: Resulting tensor has just a few nonzero columns in
each lateral slice.
Experimentally, PARAFAC seems to overfit such data and not do a good job of
“mixing” different authors.
Tamara G. Kolda – TRICAP – June 6, 2006 - p.24
Idea: Use Tucker Transformation to
Compress
or, equivalently
We transform the tensor to a smaller tensor as follows:
This transformation forces the authors to be mixed and produces a dense result.Main problem: How to transform sparse tensor without creating dense intermediate results?
(rank 75) (rank 50)
Compute rank-25 PARAFAC on compressed tensor and transform.
Tamara G. Kolda – TRICAP – June 6, 2006 - p.25
Tucker & PARAFAC
• Want PARAFAC for X in term x doc x author space
• First, apply dimensionality reduction to X to obtain Y Y in “conceptual” space
• Next, compute PARAFAC on Y
• Finally, reassemble results to yield PARAFAC for X
Tamara G. Kolda – TRICAP – June 6, 2006 - p.26
Three-Way Fingerprints
• Each of the Terms, Docs, and Authors has a rank-k (k=25) fingerprint from the PARAFAC approximation
• All items can be directly compared in “concept space”
• Thus, we can compare any of the following Term-Term Doc-Doc Term-Doc Author-Author Author-Term Author-Doc
• The fingerprints can be used as inputs for clustering, classification, etc.
Tamara G. Kolda – TRICAP – June 6, 2006 - p.27
Sample Results: Term
Tamara G. Kolda – TRICAP – June 6, 2006 - p.28
Sample Results: Term
Tamara G. Kolda – TRICAP – June 6, 2006 - p.29
Sample Results: Author
Tamara G. Kolda – TRICAP – June 6, 2006 - p.30
Summary & Future Work
• PARAFAC provides a technique for analyzing semantic graphs Third dimension captures different connection types
Or may consider it as the interconnection of 3 different node types
• Analyzed journal articles using different tensor representations Doc x Doc x Connection
• Need to make definitive case of why 3D is better than 2D
Term x Doc x Author• Too sparse?
• Still working towards large-scale, sparse problems Need implicit compression for PARAFAC
~5M nonzeros
• Other decompositions? Other hybrids
Symmetry
Tamara G. Kolda – TRICAP – June 6, 2006 - p.31
Acknowledgments & More Information
• Thanks to… Brett Bader, Danny Dunlavy, Philip Kegelmeyer Web data: Joe Kenny, Travis Bauer et al., Ken Kolda Journal data: Kevin Boyack Graph viz: Ann Yoshimura
• Related papers Algorithm xxx: MATLAB Tensor Classes for Fast Algorithm Prototyping (with B.W. Bader), ACM
TOMS, to appear. Multilinear algebra for analyzing data with multiple linkages (with D. Dunlavy and W. P.
Kegelmeyer), Technical Report SAND2006-2079, Apr. 2006. Temporal analysis of social networks using three-way DEDICOM (with B.W. Bader and
R.Harshman), Technical Report SAND2006-2161, Apr. 2006. Multilinear operators for higher-order decompositions. Technical Report SAND2006-2081, Apr.
2006. The TOPHITS model for higher-order web link analysis (with B. Bader), in Proc. Workshop on Link
Analysis, Counterterrorism and Security, SDM06, Apr. 2006 Higher-order web link analysis using multilinear algebra (with B.W.Bader), ICDM 2005, pp. 242–
249, Nov. 2005.
• Contact Info: [email protected] http://csmr.ca.sandia.gov/~tgkolda/
Thank You!