ML Journal Club
The 10-915 Machine Learning Journal Club course is designed to give students experience in making presentations. In addition, students satisfy the following MLD requirements: the Speaking Skills requirement and the Data Analysis Project (DAP) oral requirement. The course is worth 6 units. Attendance is required for those registered for credit.
Who should enroll?
- All incoming PhD students in the Machine Learning Department should register in their first or second semester. All MLD PhD students are required to take the journal club twice: typically once in their first year, during the spring semester and once in their second or third year.
- All incoming MS students in the Machine Learning Department should register in their first semester. If there is a course conflict, you may ask to defer to the second semester.
- Secondary MS students do not have to register (but are permitted to register); they should contact the instructor before the semester starts to reserve a date to give their DAP oral presentation.
- Students from other department are welcome, if space is available.
- First-time MLD PhD & MS students: must sign up for a 5-minute talk and a 30-minute talk. PhD students may sign up for a speaking skills talk as well. The speaking skills talk should be scheduled after the 30-minute talk.
- Repeat MLD PhD students: must sign up for a 5-minute talk and at least one of speaking skills (if not already complete) 30 minute, or DAP oral.
- MLD MS students: must reserve a slot in the semester when they intend to give their DAP oral.
- Students from other departments: should sign up for a 5-minute and a 30-minute, and other talks as desired and as slots are available.
Please sign up to reserve talk slots early to avoid surprises at the end of the semester.
Attendance is required for credit: part of the course is how to be a good listener as well as a good presenter. But, if you need to miss a class or two due to travel, illness, or a similar reason, it's probably OK as long as you let the instructor know before the class in question.
Each class is long enough to hold:
- One DAP oral, or
- Two 30-minute talks (including speaking skills talks), or
- One 30-minute (or speaking skills) and three 5-minute talks, or
- Four 5-minute talks
In addition, MLD rules allow at most one DAP oral per week. So, if one is already scheduled for a Tuesday, please do not request to schedule one for the same week's Thursday, and vice versa. There always seems to be a rush at the end of the semester, so please be sure to reserve your DAP slots early.
See REQUIREMENTS to make sure you schedule the right set of talks.
To schedule a talk, please follow the instructions given each semester from the instructor for the course to reserve talk slots.
In the case of a DAP oral, please also:
- get approval of the date from your advisor (advisors are required to attend DAP orals)
- make sure to provide a draft DAP document to your DAP committee for approval one month ahead of time
- send an abstract to Diane Stidle at least one week ahead of time
- check your student handbook for requirements.
In the case of a Speaking Skills talk, please also:
- check the schedule to make sure the student speaking skills committee members are available
- make sure that at least two faculty can attend and fill out your evaluation form. (Typically one of these is a Journal Club instructor and the other is your advisor, but we can accommodate other arrangements. If there are two instructors for Journal Club in some semester, do not assume that both will be present on any given day.)
- check your student handbook for requirements.
This is the “elevator pitch” talk, in which you attempt to convince the audience quickly of the importance of some research idea. (It does not have to be your own idea.) This talk should use only a very small number of slides; one of the most common mistakes is to try to include too much information. In addition, the talk should have a clear goal in mind; at the end of the talk, the audience should know what this goal is and whether they agree with it.
This is intended to be practice for a typical conference presentation, although it is longer than many. You should plan on speaking for about 25 minutes in a 30-minute slot. The talk should cover an idea in some technical detail, and be accessible to a general ML audience. It should clearly present the idea in the context of related work. Again, the idea may be yours or may be one that you have read about; in the latter case, it can be better to synthesize a few related papers, rather than depending on just one, since single papers often do not give adequate perspective on an idea.
The talk is like an ordinary 30-minute talk, except that members of the speaking committee (often including the student's advisor) attend and evaluate the presentation, as well as provide oral and written feedback to the student. See the evaluation form for an idea of what the evaluation is based on.
DAP (Data Analysis Project) Oral
The DAP oral is the culmination of a student's Data Analysis Project, like a Masters' defense, but less formal. There are a limited number of dates available for DAP orals—at most one per week.
Some common questions about DAP orals:
- How long is the talk supposed to be?
Usually the talk itself is ~45 min; together with time for questions, it will take an hour or so.
- What is the appropriate level of technical detail?
The DAP talk can and should have technical detail. Remember, though, that your audience is mostly early-year ML students; so, you'll still need to make the talk self-contained and suitable for a general machine learning audience. (I.e., it should provide sufficient background for a non-domain-expert to understand the results, and should adequately summarize the relationship of the project to previous work.) You can have a few slides accessible only to experts, but you should make sure that you don't lose the rest of the audience during these slides.
- Who needs to attend?
Of the DAP committee, only the student's DAP advisor (and the course instructor) must attend. Other members are invited but not required.
- What are the evaluation criteria?
At a basic level, there needs to be data, and an analysis of this data. The analysis methodology must be technically correct and nontrivial, demonstrating skill of the level expected from students who've taken our core courses. The talk should be well organized, well delivered, and understandable. While it is not a requirement, the DAP may include novel research, and if so, the talk should convince the audience that the research is clever and relevant.
- What are the requirements for the DAP paper?
The DAP paper is a stand-alone, single author or lead author written paper, approved by the faculty member(s) advising the DAP. It should be of high quality in exposition, English, and organization. It does not have to be published, but it should be suitable for submission to a journal or refereed conference. Unlike some conference papers, it should be completely self-contained, including all descriptions necessary for a general machine learning audience to follow the theoretical development and reproduce the experimental results. This requirement may (but does not have to) result in the project paper being substantially longer than a conference proceedings paper on which it is based.
If for some reason you have to cancel a scheduled talk, please make sure that everyone involved knows as early as possible: Diane, the instructor(s), and the speaking committee and attending faculty members if relevant. We expect, however, that students will make every reasonable effort not to cancel talks once scheduled.
We're open to suggestions, for anyone who wants to practice a particular type of speaking; for example, we've had students practice whiteboard-style lecturing, or talking without slides. The “other” talk type would typically replace a 30-minute talk.