Tuesday, 12 April 2016, 16:00 (note the unusual weekday and time)
Cybernetica Bldg (Akadeemia tee 21), room B101
Slides from the talk [pdf]
Abstract: In recent years, there has been renewed interest on column-oriented database systems, which abandon row-storage in favour of keeping the attributes of tables in separate files. Traditional vendors of row-store systems (e.g., Oracle, Microsoft) have added column-oriented features to their product lineups.
In this talk, we will propose (typed) linear algebra as a theoretical framework able to smoothly accommodate the semantics of columnar storage, coping with their quantitative and qualitative aspects in a unified way. By relying chiefly on blocked, 'divide & conquer' matrix algebra, the approach offers a framework for interpreting queries in a data-distributed, map-reduce style.