Some Subjective Slides on CS Ph.D. Admissions

Report
Some Subjective Slides on CS Ph.D. Admissions
Pieter Abbeel
UC Berkeley EECS
Why a PhD?

Become one of the world’s experts in a topic you really care
about

Technically deep and demanding

Crudely: develop new tools/techniques rather than use
existing tools/techniques
Outline

Schools

Applications

Fellowships
Schools

US News rankings are a great starting point

Pick a bunch of schools at the best level you think you can get
into, and then some safety lower-ranked schools

Make sure the schools you apply to have at least one professor
you’d love to work with

Professor who knows you well can help you get calibrated on
where you might get in and might have advice on who are the
great advisors at other schools
Why do the rankings matter?

In short, because everyone thinks they matter

In long,

Students tend to go to the highest ranked school they get into

Faculty tend to go the highest ranked school they can be at

Collaborating with and being surrounded by stronger students and
faculty tends to make you achieve more
But don’t take them too literally

E.g., potential advisor match is much more important than,
let’s say, #6 versus #8
Outline

Schools

Applications

Fellowships
What tends to get you admitted?
In short, things that stand out and are well-calibrated

Published paper or publishable work in progress

Functioning in a lab at comparable level as PhD students

At Berkeley, functioning at the level of our PhD students means functioning at
the level of #1 ranked schools’ PhD students

One of top student(s) in Ph.D. level class

Strong transcript
Some of this info will be communicated through your SoP and your letters
PS: don’t do horribly on the GREs (quantitative section carries more weight)
How about the letters?


Common advice: “Letter writers need to know you well”

Yes, at least one of them, and ideally more; in practice very few Berkeley admits have more
than one letter writer who knows them really well (this is an artifact of needing to work in a
research group for a long time to really contribute)

Potential letter writers: Research advisor, TA’ed for professor, high grade in grad course, high
grade in undergrad course
Letter writers need to be calibrated

If applying to a top ranked PhD program – how many such PhD students has that letter writer
worked with? How many undergrads have they graduated to such programs?

Does the letter writer publish papers at the same venues as the group you want to join?
If both answers are negative, the letter is going to carry very little positive weight, if any, as there
will be very limited calibrated information in there…
Letters (ctd)

Try to ask at least 6 weeks before the deadline

Don’t be afraid to send friendly reminders, e.g., 2 weeks
before, 1 week before, 2 days before, day off, 2 hours before
(hopefully no further reminders needed  )

Make sure your letter writers are happy to write you a letter
(someone feeling forced into writing you a letter might not be
writing that great a letter …)
Statement of Purpose (SoP)

Your chance to express your value to the admissions committee

2/3 what you have achieved already


Show you understand the research agenda in your lab

Explain specific sub-project you worked on, what you contributed, what was hard and
interesting about it

While research is most important, also include teaching/mentoring experience, HKN
service, outreach..
1/3 what you want to do going forward – this part is to some extent school specific,
e.g., who would you like to work with, why / on what topics? Rule of thumb,
reference 2-5 professors you’d be interested in working with, be concrete about
why; if you make it to the final rounds of the process, those professors will likely be
asked for their opinion; if you list no-one, who will they ask to vouch for you… if you
list too many, no-one listed might think you listed them all that seriously…
Your webpage

You can keep updating this even beyond the deadline


E.g., can post a paper that gets accepted in early January
Very subjective take, but stylistically either:

Super-cool website, very impressively put together

Really simple page with relevant info but that’s it because too busy
with research to embellish a web page

[don’t want to be in-between, where it looks like you tried to make a
nice page but weren’t capable of doing so …]
Get other people’s materials and feedback

Ask around in your research group

Who is willing to share their application materials?

Who is willing to give feedback on your drafts?
Fellowships

NSF and NDSEG are most prevalent ones, more or less must
apply to

Prestigious, good for resume

Really nice for your future advisor not to have to deal with fundraising for you – more time to spend with you on actual
research/advising!
NSF Fellowship

Two ratings:

Academic quality


Based on your transcript, letters commenting on your qualities, your research
proposal
Broader impacts

Outreach, tutoring, how you are changing the world (potentially also through
the research you propose)

Broader impacts is often the distinguishing factor between two
applicants who are maxed out on the academic quality score

POLISH YOUR APPLICATION, GET FEEDBACK!!
Additional Readings

http://graddecision.org/

http://www.cs.cmu.edu/~harchol/gradschooltalk.pdf
Thank you

similar documents