Applying to PhD Programs in CS

25 Jan, 2025

Notes I’m taking while I’m reading Prof Mor Harchol-Balter’s advice for grad school applications.

Future reading:

• Karpathy’s blog on a PhD: A Survival Guide to a PhD
• Matt Might's: An Illustrated Guide to a PhD

Sections:

1. Do I really want a PhD?
2. The Application Process
3. Fellowship Info
4. Choosing the right PhD programme
5. 2014 PhD programme ranks.

Do I really want a PhD?

What is a PhD?

• 6 year, in-depth (you will be the world expert or close to it in your area) exploration of a particular topic
• You know more than your advisor about your topic. You will spend the last 2 years on ONE narrow problem.

Lack of Emphasis on Courses

• MS / BA → Breadth (~3-4 courses each sem)
• PhD → Depth (@ CMU, less than 10 courses the entire 6 years)
  • Focus is Research. Take classes only if it’ll benefit in his/her research

The research process & advisor/advisee relationships

• You typically begin your research when you pick an advisor.
  • Most schools you pick one post your 1st year
  • @ CMU, you pick one post your 1st couple of months
• Usually only 50% of the admitted class graduate with a PhD. The rest can’t make the transition
• Differences b/w classes & research
  • Things are unknown in a PhD
    • During Classes, all homework problems have known answers. techniques required to solve them have also most likely been covered in class.
    • During Research, you may work for years on a problem and not know if can be solved or not. You will be the one inventing the techniques to solve this problem
  • You Choose the Problems you work on
    • During Classes, you’re assigned problems to work on
    • However in Research
      • you pick the problems you want to work on
      • you should pick good (read: “fundamental”) problems to work on.
      • you are responsible for making sure no one else has solved this problem (reading 100s of papers)
  • You’re on your own. Be Independent.
    • In classes you can reach-out and ask for help (eg classmates, Prof)
    • In Research, you’re alone. Or at best, working with your Advisor and maybe one other student. There’s no one to ask for help, because if there was someone who knew the answer, then it wouldn’t be research at all!
  • You’re your own feedback loop. No Grades, No Roadmap.
    • In classes a feedback loop exists (grades, being told what to do next)
    • In Research there is only some instruction from your advisor. The rest is up to you to be self-motivated & proactive.
  • You and your advisor become equals
    • During classes there’s a distance between you and the Prof.
      • The Prof may be dull / dry
      • The Prof has already solved the problems & just discusses results.
    • In Research, you will work side by side with the Prof
      • The dry “teaching” Prof, lights up this time
      • Prof will still
        • give you ideas for problems to work on
        • assign papers to read
        • provide a time-line / schedule
      • But when you’re working on a problem together, you will learn from each-other, make discoveries together
      • You get to see your advisor tackle problems and think out load and learn how they reason.
      • For eg:
        • You may think faster, but your advisor may have better ideas (vision)
        • You may code / compute faster but your advisor may be better @ proofs & writing / speaking
      • Many people expect their advisors to be better than them @ everything. This is unrealistic.
      • As in life, work hard @ picking up all skills that your advisor has that you don’t.
  • You will meet your advisor alone. Maximise the utility of that time.
    • During classes you rarely see your Prof alone.
    • During research you will meet your Prof ~ 1hr per week. If newer Prof then maybe ~2hr per week. It’s your job to maximise the utility of that time.
    • Profs generally
      • research
      • apply for grants
      • serve on committees
      • fly around giving talks
• All this being said, the only way to learn what research is and whether you like it is to simply start doing it.

The Application Process

• Previous Research Experience (possibly in the industry)
• Awards and Extracurricular Activities
• Transcript (Grades + Classes)
• GRE & TOEFL Scores
• Personal Statement
• Recommendation Letters

Q: Which of these are the most important?

Answer:

• Top PhD programmes:

  • Most important is the prior research and what the Reccomendation letters and Personal Statement have to say about the your Prior Research Experience
  • 3 faculty scrutinize your application, each trying to gauge your research potential

• Lesser ranked Schools:

  Simple formula probably that uses 3-ish components to bucket applications into “highly-desirable”, “less-desirable” etc.

  • Looks @ GRE
  • GPA
  • Undergraduate School