How to Grad "Student"

note

This guide applies to both MSc and PhD students. MSc is like the first year of a PhD; no expectation of being an independent researcher upon graduation.

Student vs. Apprentice

• Instead of "study" and "student", identify your work and yourself as "apprenticeship" and "apprentice".
  • Like any other apprenticeship, it's a professional endeavor.
  • The goal is to be a "master/maestro/Meister" yourself.
  • Required Reading: (Agustinus has copies.)
    • Mastery by Robert Greene.
    • So Good They Can't Ignore You by Cal Newport.
• Think of grad apprenticeship as a full-time 9-5 job.
  • The easiest, walk-in-the-park 9-5 job you'll ever do.
  • (9-5 is flexible, can also be 10-6, etc., but the point is that you have allocated time for work.)
  • See work policy of our group.
• No more undergrad lifestyle.
  • No more entire summer off
  • No more waiting to be told
  • No more treating research as an assignment
  • Etc.
• This is the baseline expectation.

Warning Signs

• Waiting to be assigned tasks.
• Treating research like problem sets.
• Not exploring literature enough.
• Working in isolation, never coming to the office, and never having in-person discussions.

Green Flags

• Initiative.
  • Example 1: Reading a new paper but confused by a concept? Find a textbook and similar papers, learn it yourself, instead of waiting for the next weekly meeting to get explanation.
  • Example 2: All tasks done? Assign new tasks for yourself.
  • Example 3: Not comfortable with some deep learning library? Find tutorials, try coding a models, etc.
• Whiteboard discussion with peers.
• Reliability: Saying that you will finish X by Y, actually finish X before Y.
• Coachability: You follow advices, e.g., during meetings, this wiki (like this page, writing advice, etc.).

Effort Scaling

• Depending on your goals, the above can be scaled up if you want.
  • E.g., if you really want to be a prof or excellent scientist, make research and science your identity, not just a job.
    • Most of the time, think of science and research. Hours blurred together between research and leisure.
    • Doesn't mean you are "switched on" all your waking hours, but you just need to observe the world during your leisure time from a scientific point of view.
    • Many "lightbulb moments 💡" occurred during leisure.
• Agustinus is more than happy to match your extra effort.
  • E.g., more whiteboard discussions, writing more papers, more guidance in learning new things (math, engineering), etc.
  • But, only if you show increased effort in the first place.
• Don't feel bad if you don't want to.
  • Agustinus is still committed to you.
    • E.g., guaranteed weekly meeting, timely feedback, etc.
  • No penalty/punishment.

Agustinus' Grad Experience

• European-style MSc: 3 semesters of coursework (around 4 courses each), 1 semester thesis.
  • Work part-time (since it's course-based and no stipend) as a research assistant, 2 (not top-tier) conference papers, 1 workshop paper.
  • Took math, theory, and ML courses; 3.9+ GPA.
• PhD: 3-year contract, no coursework.
  • Finished in 3.5 years (last .5 years were wrapping up thesis).
  • 8 conference papers (ICML, NeurIPS, AISTATS, UAI). 6 first-author papers.
  • Not once did he pull an all-nighter for a conference deadline. Low stress given how many papers he wrote!
  • Magna cum laude, won Germany-wide Best PhD Thesis Award, great postdoc placement.
• He doesn't have a superpower, nor is he a genius. He just followed his own advice and thus was "lucky".
• Probability of being lucky is increased by being disciplined, strategizing, and having good prioritization.
  • In other words, the advice above (and everything he tells you) is to increase your own "surface area of luck".

Suggested Workflow

Your working day should roughly consist of:

1. Keeping up with literature
   • Use Scholar Inbox to get daily list of reading.
   • Read in the morning when you're fresh.
   • Read the abstract, then intro, then skim the method and results.
     • Only if you think very relevant & interesting do you commit to reading in depth.
     • Remember, there is a lot of noise, and it's easy to get lost.
2. Learning fundamentals
   • Continuously advance your fundamental knowledge.
   • E.g., read one or two subsections from a textbook, do 1 math exercise, try implementing a model in PyTorch/JAX, try new data analytics & visualization technique.
   • Doing coursework counts here.
3. Advancing your research
   • Write every day, even if just a single paragraph. See: writing advice.
   • Coding, experiments, engineering, proving things count here.
   • Writing proposal, too.
4. Misc.
   • Necessary but not important for your research.
   • E.g., TA duty, group admin.
   • Should be a small portion of your time.
   • Don't let this eat away at your time for your core duties (1--3).

tip

Be very structured with your work time. Use a calendar app with an alert for everything.

Avoiding Burnout

• Burnout = doing things too hard for too long.
  • Like running a marathon with your 5k pace.
  • Or like lifting 95% of your 1RM 4x per week for 6 months.
• Smarter:
  • Each day, do the bare minimum of your duties in the previous section. E.g.:
    • Read/skim 3 papers.
    • Read 1 subsection of a textbook and solve 1 exercise.
    • Write 1 paragraph in your proposal/paper.
      note

      This will ensure you feel accomplished, make progress, and not feel guilty even when you don't feel like working.

  • Then, if you don't feel like continuing, feel free to just do light tasks like group admin or reading fun books (like Mastery and Newport's) or watching tech tutorials.
    • Or, simply end your day.
    • Go for a run, swim, or lift weights.
      warning

      Try not do only the bare minimum for too much, though! You need to make good progress in one of the focus areas above. Otherwise, you won't finish your paper, etc. Bad for your future.

  • Otherwise, continue on your current focus area.
    • You shouldn't do 100% on all three areas above at a given time. This is a recipe for burnout.
    • A better idea is to periodize.
    • Example 1: You're a new student (first semester). Then you should focus on doing coursework, reading textbooks, and reading key papers in your topic. Do the easy tasks on research, e.g., just set up environments and familiarize with tools, but do not do a lot of experiments.
    • Example 2: You've just finished submitting a paper. Then maybe do the bare minimum in advancing research, and put more time into exploratory reading.
  • Take days off every several months.
    • E.g., have a long weekend after each term, an extended vacation in the summer. See policy.

tip

Agustinus is a believer in the old saying of mens sana in corpore sano: inside a healthy body resides a healthy mind. Going to the gym or for a run in the middle of the day? Feel free to do so! This is a valid excuse to decline meeting with Agustinus or to be unavailable from your office 😄