Mathematical Models of Information Technology seminar on ‘ Inductive Databases: a Long-term Objective or a False Good Idea? ’

On Wednesday, 27 March, 2013, a regular Mathematical Models of Information Technology seminar will take place at the HSE. The event has been organized by the School of Applied Mathematics and Information Science and Intellectual Systems and Structural Analysis research and educational group, headed by Sergey Kuznetsov.

Prof. Jean-Francois Boulicaut, INSA Lyon (France) will speak on ‘Inductive Databases: a Long-term Objective or a False Good Idea?’

Abstract: One of candidate framework for the formalization of knowledge discovery processes from data (KDD) is the inductive database perspective suggested by Imielinski and Mannila in 1996 (CACM). We have been working on this database perspective on KDD processes considered as sequences of queries. The data mining steps are themselves specified by means of inductive queries and computing patterns or models is performed thanks to solvers (algorithms that evaluate more or less exactly inductive queries). We introduce this vision thanks to more or less simple examples of pattern domains: considering formal concepts in large noisy Boolean matrices or their counterpart in arbitrary n-ary relations, looking for sequential patterns in collections of event sequences or descriptive rules in n-ary relations, discovering graph patterns in (collections of) graphs, but also clustering various types of data or learning classifiers. In this talk, we propose (a) to discuss the design of pattern domains and thus the declarative semantics of inductive queries, (b) to consider solver design issues and specifically the scalability problem when looking for complete and correct though generic solvers, and (c) to see whether this has given rise to usable methods and tools for practitionners.

Some relevant references

• Rosa Meo, Pier-Luca Lanzi, and Mika Klemettinen (Eds.) Database support for Data Mining Applications - Discovering Knowledge with Inductive Queries, Springer LNCS 2682, 2004.
• Jean-Francois  Boulicaut, Luc De Raedt, Heikki Mannila (Eds.) Constraint-based Mining and Inductive Databases, Springer LNCS 3848, 2005.
• Saso Dzeroski, Bart Goethals, and Pance Panov (Eds) Inductive Databases and Constraint-Based Data Mining, Springer, 2010.

Time: 6.15 pm

Address: Room G-313 (Г-313), 11 Pokrovskiy Bulvar

Language: English,

All HSE staff and students are welcome to attend!