Gemo Seminar November 27, 2009, 14:00, room G008

Abstract

In previous work, we defined alpha Galois Lattices (GL) allowing us to summarize large data sets.
Alpha GL use a different (and coarser) extensional map according to a prior partition of data.

I will present the following work in progress related to this alpha framework :

• Incremental construction of alpha GL
• Study of Alpha rules (rules induced by the closure of alpha nodes)
• links between supervised and non supervised learning
• less-alpha extent and default logic
• clustering algorithm evaluation.

Abstract

The view selection problem consists of choosing, given a set of queries, a set of views
to materialize, in order to minimize the total cost of evaluating these queries and of
maintaining the views. Several variants of the problem exist, encompassing the existence
of a possible space constraint concerning the total size of the materialized views, or
giving weights to each query, etc. This particular problem has been thoroughly examined
for relational databases.

The increasing popularity of the RDF data model in Semantic Web and Web 2.0 applications,
has turned the efficient evaluation of queries over large volumes of RDF data of paramount
importance.

In this work, we address the view selection problem over large amounts of RDF data, given a
set of conjunctive queries over this data. We characterize the search space associated to
this problem and define a set of transformation rules over the states. Various algorithms
are presented for searching the search space.