Course organization

I. Short reports
II. Presentations
III. How to run a seminar class
IV. General Paper Commentary Outline

In the beginning of the course, students will sign up for presentations in pairs. The pair will receive a single mark on the quality of the presentation. Students who do not present on a particular day must hand in a short report summarizing the readings associated with the presentation.

I. Short reports
For each presentation, each student is required to show that they are prepared for the discussion by handing in a written one-page report for each class, critically summarizing the readings for the class.
A short report should have three distinct and clearly labelled parts:

1. Description of problem. This should provide a short statement describing the specific problem (as opposed to the general class of problems) addressed by the readings.
2. The core idea. A short description of the core technical idea, the essential contribution of the readings.
3. Discussion. A short summary of the technical issues not addressed by the readings, drawbacks or problems with the approaches, ideas on how to fix these problems or extend the approaches to solve a more general problem.

Given that this is a technical course, the focus of all short reports (and presentations) should be on the technical content of the reading in connection to the course subject, as opposed to its wider applications, philosophical underpinnings, relation with other fields, or underlying cosmic truths.
In case elaborate mathematics is used in the readings, treat it as a "black box", explaining what it does, but not how it does it. Expect that there will be parts of the readings that you don't understand. Read on and close in onto these parts by understanding all around them. If the reading were part of a thesis or project, you would normally look up other references or contact the authors to complete your understanding.

A report that avoids the technical core of the reading, but talks about "peripheral" matters at a qualitative level only will be treated as a sign of weak understanding of the core idea. A strong report must be easy to read, coherent, fluid, and must convey the key idea in the paper to someone with general background in the area but who has not read the specific paper.

Part 2 should be half the length of your report, with parts 1 and 3 each being one quarter the length of your report. The objective of the report is to give the reader a concise view of the readings and what is important about them. An effective short report is more difficult to write than a long report. Choosing what to include a very important part of writing a short report. Be highly selective. Use your own words when writing the short report. And don't forget to run a spell checking program on it in the end. Most of the guidelines for an effective presentation (see below) also apply to the short reports.

Each report is marked on a letter scale and mapped to GPA equivalent for average calculations.
Late reports are not accepted except if accompanied by a doctor's note.

Reports are due in the beginning of the class in printed form. No reports will be accepted if they are even a few minutes late.

In the short reports, if you quote whole sentences from the paper verbatim, you absolutely MUST enclose the quoted text in quotation marks ("..."). In the future, detected quotes without quotation marks will result in a grade of F on the short report. Too many quotes in the report will be taken as a sign of weak understanding of the core idea and will be given a low mark.

Common pitfalls with reports and presentations
1. Reading reports and presentations should be self contained: they should not include undefined terminology. If it is important, define it before you use it. If it is not important, omit it from the report. Terminology that has not been used in the lectures or the main textbook should be defined in the readings, if used. This limits the amount of detail that you can include.
2. Reading reports should stay at a high level. Some reports went too deep into technical detail.
3. Reading reports should be coherent. Going too deep into technical detail usually makes the reports less coherent.

II. Presentations
Each of these presentations should be at most 20 minutes long. You should time your presentation very carefully, as the time limit will be strictly enforced.

Considerations for presentations
content
presentation style
sample case for illustration
discussion

When preparing the presentation, it is important to keep in mind the following:

1. Digest the material first. If certain things in the reading are not easily accessible, isolate them, and digest everything else first. Try to get to the core issues in the reading, the essential contributions, the philosophy of the author, why the reading is important and worth studying. Do not move to the preparation of the presentation before you understand the reading well.

2
.Take into account your audience when preparing the presentation. Wear the hat of another student in the class, who may have a different background than yourself. Don't assume too much about the audience. Students are supposed to have studied the reading, but the presenter has spent more time on it, and knows the material better.

3
. Be selective in what you include in the presentation. Too much detail is hard for the audience to follow. Not enough detail results in a talk that is not concrete enough. The big picture is more important. Use figures and diagrams and minimize text. Do not present mathematical formulas, except when they are important and you are prepared to focus on them. Do not use full sentences or paragraphs, itemize issues instead. Explain the intuition behind the mathematics.
The title page of your presentation should have the full title, authors and publication venue of the reading, followed by:
An in-class summary by <your names>.


4
. A good way to start your presentation is by walking through a simple illustrative example or sample case. If the reading has such an example, use it. If it does not, construct one of your own.

5
. Anticipate questions from the audience and be prepared to explain them in more than one way. Record your own questions the first time you studied the reading. Record your line of thought when you answered them. If the reading left
some questions unanswered, identify them, and bring them up for discussion in class. Consult the references of the reading or related readings to get answers and complete your understanding. Your ability to handle questions is a very good indicator of your understanding of the material.

6. Use of visual aids: Drafting figures and diagrams on the computer is very time consuming. Working with electronic presentation tools like PowerPoint may lead to more text and less figures than optimal. Feel free to copy and paste figures from the paper into your presentation.

7. Timing of the presentation: Management of time is very important when giving your presentation. Take only short questions during the presentation but do not engage in long discussions, as this interrupts the flow of the presentation and may consume valuable time forcing the presenter to leave material out due to lack of time. Long discussions should take place after the end of the presentation. If the presenter fails to defer long discussions for later, the instructor reserves the right to do so.

8. The timing of the material is very important. It is easy to prepare too much material and then take too long to present it. Be selective in what you include in the presentation. Use your understanding of the paper as the guide on what to include.

9 . Leave out stuff that is trivial or "operational" ("how to" details that are not relevant to the core concepts). Avoid general or encyclopedic statements that do not relate to the core concepts of the paper.

10. Where there are multiple similar algorithms for the same problem, choose one to present the concept. You do not have to be exhaustive in your coverage.

11. You do need to convey the key concepts in an intuitive manner. You must present a coherent "story" in your presentation, that captures the attention of your audience. Your audience will be attentive if they learn something new from the presentation, that does not go over their head, and does not bore them with too much detail.

12. As a guideline, for a 30 minute presentation, you should have no more than 20 slides in total, including figures, formulas, browser screens.

13. Giving good presentations is a necessary skill for advancement in the workplace. There is no talent in giving a good presentation, only hard work in understanding your material, in preparing your overheads and in rehearsing the talk enough to feel in command.

More guidelines on presentations

14. Presentations should not exceed 20 minutes. Use your watch/clock during your presentation to monitor time.
A presentation by a pair should conform to the same time limit as an individual presentation.

15. It is OK for only one member of the pair to do the whole presentation. The assumption is the two students worked together to understand the paper and to produce the visual aids.

16. The objective of the presentation is to review the key points of the paper, critique the results, and promote discussion. It should avoid too much technical detail. It should emphasize qualitative issues. At the same time, it should focus on the core idea, and avoid "fluff".

More advice on giving good presentations


III. How to run a seminar class (roles)

Organizers
Review literature and post suggested readings for the class
Organize presentation
Organize discussion format
Prepare case study or example

Participants
Read selected readings plus at least 1 other source
Prepare a 2 page report.

Instructor's Role
set report topic (sample and questions)
meet with presenters before class


IV. General Paper Commentary Outline
by Prof. Nathalie Japkowicz

What is the research goal?
What question(s) is the author trying to answer?
What methods are being applied?
What methods is the author applying to answer the question?
What are the research results?
A paper can contain many different kinds of results (E.g.: applied results (efficiency or accuracy), theoretical results (a theorem was proven), etc.)
What claims are made in the paper?
For theoretical papers, what results are proven?
How are these claims supported?
What reasonable claims and results are missing from the paper?

What would be reasonable next steps for the research?

How to initiate a paper discussion
by David Jensen, 2001 (http://kdl.cs.umass.edu/courses/csna/initiate-discussion.html)


Specific Example: NIPS Review Criteria

Quality -- Is the paper technically sound? Are claims well-supported? Is this a complete piece of work, or merely a ``position paper''? Are the authors careful (and honest) about evaluating both the strengths and weaknesses of the work.
Clarity -- Is the paper clearly written? Is it well-organized (if not, feel free to make suggestions to improve the manuscript)? Does it adequately inform the reader? (A superbly written paper provides enough information for the expert reader to reproduce the results.)
Originality -- Are the problems or approaches new? Is this a novel combination of familiar techniques? Is it clear how this work differs from previous contributions? Is related work adequately referenced?
Significance -- Are the results important? Does the paper address a difficult problem in a better way than previous research? Does it advance the state of the art? Does it provide unique data, unique conclusions on existing data, or a unique theoretical or pragmatic approach?