In this project, we explore diversity in
the context of large-scale and numeric multi-dimensional databases. We develop
query processing methods and tools for browsing and exploratory queries. Such
queries should return diverse results, each contributing to the novelty and
diversity of the result set. We propose the idea of diverse matching that
utilize subspace similarities to avoid the curse of dimensionality. Our approach
is based on optimizing the type of partial proximity each result possesses. In
particular, we maximize the coverage of attributes that are found similar in
the result set, while minimizing the overlap of these sets attributes. For
large-scale databases, we study a browsing method over metric indices based on
angular pruning.
We also have another project on
multi-dimensional big data.
Finally, our past research is here.