Information Theoretic Clustering, Co-clustering and Matrix Approximations
         Inderjit S. Dhillon
                       University of Texas, Austin

Introduction

Clustering

Co-clustering

Matrix Approximations

Slide 6

Slide 7

Co-clustering and Information Theory

Information Theory Concepts

Jensen-Shannon Divergence

Information-Theoretic Clustering: (preserving mutual information)

Information Theoretic Co-clustering (preserving mutual information)

Slide 13

Preserving Mutual Information

Example – Continued

Co-Clustering Algorithm

Properties of Co-clustering Algorithm

Slide 18

Applications -- Text Classification

Dimensionality Reduction

Experiments

Naïve Bayes with word clusters

Results (20Ng)

Results (Dmoz)

Results (Dmoz)

Slide 26

Results (Dmoz)

Example

Co-clustering Example for Text Data

Results– CLASSIC3

Results – Sparsity

Results – continued

Results (Monotonicity)

Related Work

Conclusions

Contact Information