Submitted by tkujala on September 7, 2007 - 00:00
HIIT seminars in fall 2007 will be held in hall **B222** of Exactum,
on Fridays starting at 10:15 a.m. Coffee available from 10.
Fri Sep 7
Tomi Silander
On Learning The Most Probable Bayesian Network
Abstract:
Learning the MAP Bayesian network for a complete discrete data is known
to be an NP-hard problem. Consequently, much of the research effort has
been concentrated on developing heuristics for this important task.
However, recent advances in exact methods for learning Bayesian Networks
allow us to study properties of the most probable Bayesian networks
revealing other complications besides the hard learning problem.
In this presentation, we demonstrate a considerable sensitivity to the
(usually carelessly chosen) parameter prior, offer some alternatives to
solve the sensitivity problem, and wonder if learning the most probable
Bayesian network is a good problem at all.
on Fridays starting at 10:15 a.m. Coffee available from 10.
Fri Sep 7
Tomi Silander
On Learning The Most Probable Bayesian Network
Abstract:
Learning the MAP Bayesian network for a complete discrete data is known
to be an NP-hard problem. Consequently, much of the research effort has
been concentrated on developing heuristics for this important task.
However, recent advances in exact methods for learning Bayesian Networks
allow us to study properties of the most probable Bayesian networks
revealing other complications besides the hard learning problem.
In this presentation, we demonstrate a considerable sensitivity to the
(usually carelessly chosen) parameter prior, offer some alternatives to
solve the sensitivity problem, and wonder if learning the most probable
Bayesian network is a good problem at all.
Events:
Last updated on 12 Sep 2007 by Martti Mäntylä - Page created on 7 Sep 2007 by Teija Kujala