|
|
Seminar / Practical Course "Probabilistic Logic Learning"
Prof. Dr. Luc De Raedt
Co-organizer: Dipl.-Inf. K. Kersting, Dipl.-Inf. A. Zimmermann
(Seminar) Wednesday 16-18 o'clock, Room: SR 00-019, Building 079
(Practical Course) Thursday 14-18 o'clock, Room: SR 00-019, Building 079
Credit points (Kreditpunkte): (Seminar) 3, (Practical Course) 6
- Report subsmission deadline:
March 15th, 2004. Any report submitted later will not be considered.
- Demo sessions will take place on
Thursday February 12th, 2004, 16 - 18 o'clock and
Friday 13th February, 2004, 11-13 o'clock.
 local course web page
The past few years have witnessed an significant interest in probabilistic logic learning,
i.e. in research lying at the intersection of probabilistic reasoning, logical representations,
and machine learning.
A rich variety of different formalisms and learning techniques have
been developed by many researchers from a variety of backgrounds (including machine learning, statistics,
inductive logic programming, databases, and reasoning under uncertainty). The aim of this research line
diverges from traditional approaches in these fields that assume data instances are
structurally identical and statistically independent or
assume that relationships are deterministic. Several workshops
(SRL-00, MI-19, SRL-03), reasearch projects ( EELD, APrIL I, APrIL II, ...), and (invited/honorary) talks (such as D. Koller at IJCAI-01, ICML-03/KDD-03 and F. Provost at ICML-03) have been devoted to probabilistic logic learning.
The courses will explore approaches to probabilistic logic learning.
We will explore the
- foundations (i.e. knowledge representation, reasoning and learning within both traditional and
upgraded probabilistic frameworks),
- tasks and applications (e.g. collaborative classification and filtering, link discouvery, bioinformatics, ...), and
- advantages, and limitations of
the surprising array of approaches that have been developed over the past decade.
More precisely, we will mainly focus on probabilistic-logical models (PLMs).
PLMs integrate probability theory with some first order
logic. Traditionally, a probabilistic formalism
like Bayesian networks or hidden Markov models is selected and upgraded
by incorporating some logic such as entity-relationship (ER) models, Datalog,
or Prolog. Frameworks developed include
probabilistic relational models (PRMs), stochastic logic programs (SLPs), Bayesian logic programs (BLPs),
relational Bayesian networks (RBNs), relational probability trees,
first-order Bayesian classifiers, relational Markov models,
block models, statistical relational (SRMs) models and relational reinforcement learning (RRL).
|
