Radim Jiroušek presents 58th edition of the PRAGUE COMPUTER SCIENCE SEMINAR

From Graphical to Compositional Models

The lecture will be followed by a discussion

ABSTRACT

“Bill Gates, the richest man in the world, loves Bayesian networks.” This
statement appeared in the titles of many newspaper and journal articles after
October 28, 1996, when the Los Angeles Times published an interview with him. He
said that Microsoft’s competitive advantage was its expertise in Bayesian
networks. This declaration indisputably contributed to the popularity of
Bayesian networks, and since that time, they have become a standard technique
incorporated in statistical and data-mining software packages. However, Bayesian
networks are only a part of the technology developed to handle multidimensional
probability distributions efficiently.
The lecture will therefore also mention other, less famous, graphical models, as
well as nongraphical tools, especially, compositional models. It will also show
that most of the results valid for probability may be extended to other
uncertainty theories, especially to belief functions.
Thus, the audience will learn how to differentiate between combination and
composition, and, for example, what belief functions have in common with the
furniture sold by IKEA. The talk will be an author’s subjective glimpse of
almost 40 years of history and an opportunity to recollect some names that
should be remembered.