A Graph Theoretic Approach to Heterogeneous Data Clustering

Data clustering is the process of automatically
grouping data objects into different groups
(clusters). The contribution of this book is
threefold: homogeneous clustering of images, pairwise
heterogeneous data co-clustering, and high-order
star-structured heterogeneous data co-clustering.
First, we propose a semantic-based hierarchical image
clustering framework based on multi-user feedback. By
treating each user as an independent weak classifier,
we show that
combining multi-user feedback is equivalent to the
combinations of weak independent classifiers. Second,
we present a novel graph theoretic approach to
perform pairwise heterogeneous data co-clustering. We
then propose Isoperimetric Co-clustering Algorithm, a
new method for partitioning the bipartite graph.
Lastly, for high-order heterogeneous co-clustering,
we propose the Consistent Isoperimetric High-Order
Co-clustering framework to address star-structured
co-clustering problems in which a central data type
is connected to all the other data types. We model
this kind of data using a k-partite graph and
partition it by considering it as a fusion of
multiple bipartite graphs.

Autorentext

Prof. Manjeet Rege, Ph.D. is with the Department of ComputerScience at Rochester Institute of Technology. Prof. Ming Dong,Ph.D. is with the Department of Computer Science at Wayne StateUniversity. Their research interests lie in the areas of DataMining, Machine Learning, Information Retrieval, and MultimediaContent Analysis.

Klappentext

Data clustering is the process of automaticallygrouping data objects into different groups(clusters). The contribution of this book isthreefold: homogeneous clustering of images, pairwiseheterogeneous data co-clustering, and high-orderstar-structured heterogeneous data co-clustering.First, we propose a semantic-based hierarchical imageclustering framework based on multi-user feedback. Bytreating each user as an independent weak classifier,we show thatcombining multi-user feedback is equivalent to thecombinations of weak independent classifiers. Second,we present a novel graph theoretic approach toperform pairwise heterogeneous data co-clustering. Wethen propose Isoperimetric Co-clustering Algorithm, anew method for partitioning the bipartite graph.Lastly, for high-order heterogeneous co-clustering,we propose the Consistent Isoperimetric High-OrderCo-clustering framework to address star-structuredco-clustering problems in which a central data typeis connected to all the other data types. We modelthis kind of data using a k-partite graph andpartition it by considering it as a fusion ofmultiple bipartite graphs.