Decoration
max planck institut
informatik
mpii logo Minerva of the Max Planck Society
 

Seminar "Statistical Relational Learning", Winter 2011/2012

Organization

Contents

This seminar provides an introduction to statistical relational learning (SRL). SRL attempts to represent, reason, and learn in domains with complex relational and rich probabilistic structure. Applications include named-entity recognition, coreference resoltuion, web-scale knowledge base construction, object identification in images, spam classification, and many more. The seminar closely follows the book "Introduction to Statistical Relational Learning" (The MIT Press, 2007, edited by Lise Getoor and Ben Taskar) and covers a variety of topics such as inductive logic programming, conditional random fields, probabilistic relational models, Bayesian logic programming, and Markov logic.

Goals

In this seminar, you will
  • Read, understand, and explore scientific literature
  • Summarize a current research topic in a concise report (5 pages)
  • Give a full presentation about your topic (45 minutes)
  • Give a flash presentation about your topic (5 minutes)
  • Moderate a scientific discussion about a topic of one of your fellow students

Requirements for the Certificate

  • Pick a topic from the list below. Prepare a 45 minutes presentation about your topic to introduce it to your fellow students.
  • Make a first appointment with your tutor (who will be announced along with the topics) to discuss the outline of your presentation at least 4 weeks in advance of your presentation. You are responsible for scheduling meetings with your tutor.
  • Point out advantages or potential weaknesses of the work covered in your presentation. If you are unsure about what to present, talk to your tutor. Note that—even though relevant presentations may be available on the web—we expect that you prepare your own slides (which may be, of course, inspired by the original slides). Send your slides to and discuss them with your tutor at least 2 weeks before your talk. Otherwise, your talk may be cancelled.
  • Each presentation is followed by approximately 15 minutes of discussion. The discussion is moderated by a second student. The moderator's role is to provide interesting input (such as observations, questions, related work) for the discussion and, in general, to enable a constructive discussion. A preliminary version of the presenter's slides will be sent to the moderator on the Friday before the presentation.
  • Three weeks after your talk, submit a short report (not longer than 5 pages) about your topic. The report should concisely summarize the article and point out strengths and weaknesses.
  • In our last meeting, give a 5 minutes flash presentation about your topic. As before, discuss your slides with your tutor at least 2 weeks before the presentation.
  • Attend all presentations, not just your own. If you are ill, let us know in advance.
  • Actively participate in the discussions.
  • Slides, presentations, and reports must be prepared in English.
  • Your final grade is influenced by: your oral presentations, your knowledge about your topic (e.g., as shown in the discussion after your presentation), your performance as a moderator, your general participation in the seminar, and your written report.

Tentative Schedule

  • Oct. 18, 2011: Introduction (SRL, organization, how to give presentations) (pdf)

  • Nov 29, 2011: Inductive Logic Programming in a Nutshell (Saso Dzeroski)
    Presenter: Humayun Faiz
    Moderator: Indra Praveen Sandrala
    Tutor: Christina
    Materials: slides (pptx), flash slides (pdf), report (docx)

  • Dec 6, 2011: An Introduction to Conditional Random Fields for Relational Learning (Charles Sutton, Andrew McCallum)
    Presenter: Sanjar Karaev
    Moderator: Arnab Dutta
    Tutor: Rainer
    Materials: slides (pdf), flash slides (pdf), report (pdf)

  • Dec 13, 2011: No talk

  • Dec 20, 2011: Stochastic Logic Programs: A Tutorial (Stephen Muggleton, Niels Pahlavi)
    Presenter: Farzaneh Ansari
    Moderator: Milos Ercegovcevic
    Tutor: Christina
    Materials: slides (ppt), flash slides (ppt), report (pdf)

  • Jan 10, 2012: Probabilistic Relational Models (Lise Getoor, Nir Friedman, Daphne Koller, Avi Pfeffer, Ben Taskar)
    Presenter: Ondrej Sykora
    Moderator: Humayun Faiz
    Tutor: Rainer
    Materials: slides (pdf), flash slides (pdf), report (pdf)

  • Jan 17, 2012: Bayesian Logic Programming (Kristian Kersting, Luc De Raedt)
    Presenter: Indra Praveen Sandrala
    Moderator: Sanjar Karaev
    Tutor: Rainer
    Materials: slides (pdf, pptx), flash slides (pdf, pptx), report ( pdf)

  • Jan 24, 2012: Lifted First-Order Probabilistic Inference (Rodrigo de Salvo Braz, Eyal Amir, Dan Roth)
    Presenter: Arnab Dutta
    Moderator: Farzaneh Ansari
    Tutor: Christina
    Materials: slides (pdf), flash slides (pdf), report (pdf)

  • Jan 31, 2012: Global Inference for Entity and Relation Identification via a Linear Programming Formulation (Dan Roth, Wen-tau Yih)
    Presenter: Milos Ercegovcevic
    Moderator: Ondrej Sykora
    Tutor: Rainer
    Materials: slides (ppt), flash slides (ppt), report (pdf)

  • Feb 7, 2012: Flash Talks

Literature

  • Introduction to Statistical Relational Learning, edited by Lise Getoor and Ben Taskar, The MIT Press, 2007.
  • Writing for Computer Science: The Art of Effective Communication, Justin Zobel, Springer, 2004.