Paper Presentation Guidelines
CMSC 671, Fall 2010

Selecting a Paper

• Symmetry in Solutions, Marijn Heule and Toby Walsh
• Searching Without a Heuristic: Efficient Use of Abstraction, Bradford John Larsen, Ethan Burns, Wheeler Ruml and Robert Holte
• Search Space Reduction Using Swamp Hierarchies, Nir Pochter, Aviv Zohar, Jeffrey S. Rosenschein, Ariel Felner
• A Proof-Producing CSP Solver, Michael Veksler and Ofer Strichman
• Decomposed Utility Functions and Graphical Models for Reasoning about Preferences, Ronen I. Brafman and Yagil Engel
• Ontologies and Representations of Matter, Ernest Davis
• Assisting Users with Clustering Tasks by Combining Metric Learning and Classification, Sumit Basu, Danyel Fisher, Steven M. Drucker and Hao Lu
• Learning Causal Models of Relational Domains, Marc Maier, Brian Taylor, Huseyin Oktay and David Jensen
• Multi-Task Active Learning with Output Constraints, Yi Zhang
• Nonmanipulable Randomized Tournament Selections, Alon Altman and Robert Kleinberg
• Competing Schedulers, Itai Ashlagi, Moshe Tennenholtz and Aviv Zohar
• Truth, Justice, and Cake Cutting, Yiling Chen, John Lai, David Parkes and Ariel D. Procaccia
• Cloning in Elections, Edith Elkind, Piotr Faliszewski and Arkadii Slinko
• Multi-Agent Plan Recognition: Formalization and Algorithms, Bikramjit Banerjee, Landon Kraemer and Jeremy Lyle
• Planning in Dynamic Environments: Extending HTNs with Nonlinear Continuous Effects, Matthew Molineaux, Matthew Klenk and David Aha
• Temporal Information Extraction, Xiao Ling and Daniel S. Weld

Presentation

• Convey why this is an important and/or interesting problem
• Review key ideas in the paper
• Convey why this is an important and/or interesting approach
• Identify the main contributions and claims of the authors, and how they supported those claims analytically or empirically
• Critique the work

• Content: You should provide a well organized presentation of the key contributions and important ideas in the paper.
• Timing: You should aim for a 10-minute presentation. This works out to (roughly) five to six slides---no more! As in a real talk, you will get 5-minute, 2-minute, and time's-up warnings from the session chair. I will cut you off if you go too long! There will be a minute or two after each talk for questions.
• Audience: Your audience consists of computer science graduate students. (I don't count.) They will not have read the paper, so you can't assume a lot of existing knowledge, On the other hand, you only have ten minutes! Be selective!
• Organization: Suggestions:
  1. Describe the problem
  2. Starting with a simple example can be very helpful
  3. Explain why it's important (or at least why the authors think it's important)
  4. State how the authors solved the problem at an appropriate level of detail
  5. Tell what explicit and implicit claims the authors make
  6. Describe the authors' experimental and/or analytical evidence for these claims (and indicate whether you think the evidence is sufficient to support the claims)
  7. Stimulate discussion by pointing out interesting aspects of the approach, flaws, limitations/assumptions, open questions, ...
• Feedback: Draft slides can optionally be sent to me for review, if you want feedback beforehand. (You must send me your slides at least 24 hours before your talk or I won't guarantee to look at them and give you comments.)
• Grading: Students are required to fill out a short feedback form for each presentation. You will receive these forms for your information, and I will also give you written feedback. Your grade will be based on your level of preparation, the clarity of your presentation, the timing of your presentation, and other students' evaluation of your presentation.
• Other advice: Practice your presentation, even if it's just to yourself, to make sure your timing is correct.

Strategies for Selecting and Reading Papers

• Decide from the title and context if it might be interesting/relevant.
• Read the abstract to decide if it actually is relevant, then read through the introduction quickly to make sure.
• Once you've decided to read the paper, don't just tackle the paper head-on. Approach it strategically.
  • First skim the whole paper once to get a feeling for what it's about. Skip anything dense or technical. Glance at the figures and tables.
  • Read through the whole paper, but don't let yourself get stuck.  If you have questions, don't struggle to figure them out; instead, jot a note in the margin and keep going.  Put stars next to important points and/or underline key ideas.  Don't get carried away with the underlining, or it won't be helpful when you want to look back over the paper.
  • Go back and work through the details of any (important) equations, proofs, algorithms, etc. that you skipped over.
  • You may need to look up citations of work (by the authors or by others) on which this paper builds.  It's important, though, to learn which papers are critically important and which are less so -- it would be entirely possible to spend the next year (or two or three or four) doing nothing but reading, and still not have read all of the papers that you "ought" to. (When you move into your own research project: Spend your time wisely, and keep track of your reading so you don't have to re-read everything next year.)

  Some Useful Links

  http://www.biochem.arizona.edu/classes/bioc568/papers.htm
  How to Read a Scientific Paper, John W. Little and Roy Parker, University of Arizona.
  Oriented towards biology papers, so not everything is relevant, but section 4, "Evaluating a Paper," has some particularly helpful suggestions.

  http://swig.stanford.edu/~fox/paper_writing.html
  Armando's Paper Writing and Presentations Page, Armando Fox, Stanford University.
  Mostly on writing and presenting papers, but also talks about reviewing papers.

  http://www.isi.edu/gost/courses/csci555/survive/
  CS 555: How to Survive!, John Brewer, ISI.