Adam's principles on lab meeting in graduate school
Lab Meetings
Every graduate student may have memories of lab meetings. (and may be preparing a meeting right now..) Here are some tips for successful lab meetings from my experience.
Presenter
Prepare your material according to ‘scientific method’ as much as possible!
Basic Settings
Please don’t just enumerate what you’ve done, and please stick to the time limit as if it were a conference presentation (25 minutes).
Your & our time is gold.
Human’s concentration is limited. That’s why Pomodoro technique usually suggests 25 minutes timer.
Please prepare your presentation seamlessly and carefully to make the lab meeting meaningful for everyone.
Take ownership: no one knows every detail of your work better than you.
Template
I. Introduction
Introduce a clear definition of the target problem.
Give your lab mates a quick refresher on the topic you are working on and the importance of your work.
II. Observation & Background
Introduce from where, and why you found it is really a “problem”?
Provide summaries of related works you have read & keywords used to find/define such works.
Where can I find meaningful problems?
For Master students/industry-biased) you can find a problem from industries you are interested in (e.g., see the requirements sections in job postings!)
For Doctoral students/academy-biased) you can find a problem from papers, or by communicating with other researchers (e.g., conferences/twitters/reddits), or by cloning GitHub repo. of previous works and playing with them, …
Conducting a research is like “debugging codes” with other researchers around the world.
Without knowing what the previous bug is, (and whether it has already been fixed or not), any patches/updates are pointless…
Please DON’T "imagine" a research problem… You need to cross-check if there is agreement among other researchers that the problem you found is indeed a problem. (Before making PR, you make an issue first, innit?)
Please DON'T DON'T DON'T reinvent the wheel!!!
III. Hypothesis
Provide the very point(s) that you’d like to make a progress on existing works.
Please describe in a compact sentence containing core keywords.
It becomes the main thesis of the paper
Template: [independent variable(s) / your idea] will [expected effect] on [dependent variable / target field of interest] in [specific condition / essential assumption]
Example)
[A server-side online convex optimization on mixing coefficients] will [induce better performance uniformity] on [both cross-silo & cross-device federated learning] in [a single model-based, honest server and reliable clients in federated setting] (See AAggFF (ICML`24) for more details)
IV. Objective
Summarize in a compact sentence what you have tried before a lab meeting.
Example) To investigate FTRL framework can be adopted to federated setting even with client sampling.
V. Experimental Results
Summarize your results in a sentence.
Then, answer the following questions:
How and why did you design experiments to test your hypothesis?
What do the results imply about your hypothesis?
VI. Discussion
Wrap-up your thoughts and provide future plans based on the hypothesis and corresponding results.
Request critical feedback from your labmates.
Please be specific about what you’d like to discuss.