Confessions of a researchaholic

April 15, 2018

More about choosing graduate programs

Filed under: Real — liyiwei @ 5:19 pm

Recently several people asked me about choosing graduate programs. I answered about this before, but some specific cases might still help.

Q: Choose a program that offers financial support over one that does not.

This is certainly a sensible decision, especially if you are cash strapped, even though that is not what I did (see below).

Q: Choose a MS program that is more likely to lead to the PhD program of the same school/department.

In general yes, if you can find a good adviser whom you can convince to take you as a PhD student by doing good research with him or her.

Q: Choose between a school/adviser good at a direction or a methodology

These are different. For example, you might be interested in computer graphics (direction) and want to apply machine learning (methodology) in some way. Should you choose a professor in computer graphics who knows a bit (but not much) about machine learning, or a machine learning expert who is not really doing computer graphics?

Always pick the one who is better at identifying the problems (directions) than the one who is better at suggesting solutions (methods). Solving a wrong problem (unimportant, too easy, too difficult, too crowded, etc.) can completely derail your research career. In contrast, you can find domain experts to collaborate after identifying a good direction.

[I will add more when I receive more questions.]

When I made my decision more than 20 years ago, it was pretty easy: I wanted to go to a top school in the Bay Area, so I basically had only two choices. And I heard from one around Christmas but not another one until the next April or May. What was more difficult is that I only got scholarships from schools in the east coast, and my family preferred me to go there.


  1. (Not really a question.)

    Q: What if feasible profs are good at neither direction nor methodology, for example?
    A: Find those who are good at math.

    Comment by Jiayao Zhang — April 19, 2018 @ 12:34 am | Reply

  2. I am interested in machine learning. If the supervisor is pretty new (i.e. Assistant professor who just joined the university), but good at math (he is representative of China Math Olympic when he was young), but didn’t publish any novel methodology paper on machine learning but he is interested in this area too as indicated on his website (because he was working in the industry as a statistician before after getting his PhD in statistics), how would I go about making the best research experience possible during the 4 years as a PhD (he has accepted me but my grade obviously doesn’t allow me to get into top school, so I decided to accept the offer from this professor). Should I go figure out a topic on my own by reading the website from some famous professor who is good in this field and try to identify a direction on my own (which seems very difficult as I don’t have enough knowledge about machine learning to begin with) and then try to pursue the direction as understood from this famous professor website (and ask my supervisor for help during the time as a PhD), or should I just do my supervisor’s project which is also about machine learning but might not be on the methodology development side.

    Comment by pak gor wong — May 8, 2018 @ 11:12 pm | Reply

RSS feed for comments on this post. TrackBack URI

Leave a Reply

Your email address will not be published. Required fields are marked *


This site uses Akismet to reduce spam. Learn how your comment data is processed.

Theme: Rubric. Get a free blog at