Advice for 1st year PhD students @ DTIC UPF

Who am I?

Home page: https://yipingnus.github.io/

I completed my bachelor’s, master’s, and PhD in Southeast Asia. I moved to Barcelona in 2022 to study an intensive Spanish course. After spending nearly a year as a visiting postdoc at UPF, I started doing a PhD here again. But why?

Of course, a PhD degree is the entry ticket to many academic jobs. But it’s beyond that. It’s 3 years of uninterrupted time you dedicate to become an (ideally world-class) expert in a field. I completed my first PhD while working full-time in a company. My publications were alright, but they were scattered across different topics and I never had the luxury to dive deep enough due to the short time span of the projects. I’m fortunate enough to have sufficient financial resources and autonomy. That’s why I considered devoting another 3 years to research and see how far I can get.

The advices I give in this page are ones I wish I would have heard when I started my PhD. It’s given by both someone who has gone through it and someone who is going through it right now. My hypothesis is that most people don’t know how to finish a PhD (well), even those who have finished it. They might get lucky and get a few papers accepted. Before they have the time to reflect, they become Dr. XX and have to move on apply for jobs. Young people don’t like to listen to advices. But bear in mind: It’s much more costly to learn from our own mistakes than learning from other people’s mistakes.

A PhD is not enough. A book that I can’t recommend enough. You can get the free PDF version.

#1: It’s your PhD. Not your supervisors

Your supervisors are experienced researchers. They have developed the research taste discerning which idea is promising. They can give you pointers to help you publish faster. However, that research taste is something you want to develop. If you only do what your professor tells you to do, you will waste your PhD. You need trail and error and the grind to learn and grow as a scientist.

Your autonomy depends on

Funding. If you’re supported by a project, you’ll have much less freedom (but still more than you thought).
Supervisor’s personality. Authoritarian vs. liberal.
Your track record. If you consistently deliver results, you’ll win yourself autonomy along the time.

#2: Better sitting on a bench in premier league than being a superstar among amateurs

The competitive structure in CS research is extremely strong. You should aim for the very top conferences instead of lowering your standard and submitting to any conference that’s not in the top 5 list. I’m surprised that some students started their PhD without knowing which are the top conferences they should submit to (and read). If it’s the case for you, google “google scholar top conferences [YOUR FIELD]”. The h-index isn’t indicative, but the top 5 conferences are more or less accurate.

Top conferences in NLP

Of course, this means you should get used to being rejected (often). Even if your paper is rejected, the negative reviews will help you improve your paper and skills (and your character) much more than positive ones.

#3: Keep your ego in check and keep doing hard work

The reward signals in PhD are extremely sparse:

A negative reward (paper rejection) can make you feel it’s the end of the world.
A positive reward (paper acceptance) also generates a lot of paperwork (camera-ready, releasing code, preparing talk, attending conference, making claims) and excessive excitement.

Either of these can knock you out of your productivity for weeks to months. My advice is after wrapping up a (successful/unsuccessful) project, take a brief moment for appreciation and recharge, then move on to the next one, then the next one.

I’m from a very competitive culture where we establish our worth by outcompeting our peers. For each of your new paper, do the appropriate publicity (conference talk, invited seminar, Twitter, blog …). Then forget about it and leave for the people to discover it. If your work is really good, people will find it sooner or later. Don’t fall into the trap of monitoring your citation counts like the stock market.

#4: Follow the gradient

This is an important advice I would give my past self: Build your papers on top of the previous ones instead of shooting in random directions.

The overhead for you to produce elite result in a new sub-field is very high.
It significantly reduces the effort producing each paper.
It helps you build a cohesive story line for your thesis.

#5: Take care of your hardware

So far, I might seem overly harsh (and realistic). My final advice is that you should take care of yourself.

A great workday starts from a good night of sleep, lifting heavy weights in the morning, an optimized nutrition plan.
If you have problem with your physical/mental health, you’re making an already hard job much harder. Why do so?
In the long run, PhD is a very “easy” job compared to anyone you’ll have after completing your PhD.

That said, I wish you a successful and happy PhD journey!

Keep in touch:

My office is at 55.420. Feel free to walk in and chat if you have any question.

Or drop me an email [first name].[last name]@upf.edu.

Yiping's Blog