On choosing the right PhD for you

I am one of those lucky individuals doing research in one of the most happening fields in the world. The field that is often seen as the “electricity of the modern era”. Yes, I work in Machine Learning (ML). I work on its application to understand images and videos — Computer Vision. Though my perspective in this story is biased towards ML, it applies to most PhDs.

ML is a very active research field these days. There are billions of dollars floating around in terms of funding. Hundreds of academics are being approached by companies for consulting. For graduating Masters students this means ample opportunities for doing a PhD. From my experience of interacting with ~5 years of graduates from MSc programme, I notice a common trait: Top students get flooded with PhD offers and simply do not know how to choose the right PhD catered for them. As a result they regret having wasted their valuable years doing a PhD. Mediocre students just flow with the tide and go for the PhD that comes their way and regret having wasted their time. So, I wish to unleash some unwritten pointers for choosing the right PhD that suits ‘you’.

Topic

A PhD is a constant grind to explore the unknown aiming to push the boundary of a given tiny area of a field. As Andrej Karpathy puts it in his blog, “A PhD is simultaneously a fun and frustrating experience”. So to get a good PhD in the end (in n years 😃), it is better to choose a topic in an area of your field that is extremely interesting for you, excites you or has always been exciting. It is quite important not to be tempted by any other material temptations such as:

• A prestigious scholarship. Most students accept these scholarships thinking that they make the CV look good. Not to forget that these scholarships demand a lot from the student: strict direction, meeting deadlines, making reports, a supervisor who is not fully academic, etc. As a result you lose the joy of doing research. In my opinion, the papers you publish during your PhD are much more valuable than just winning a prestigious scholarship to get you into a PhD. Your papers can bring recognition and the positions you get after your PhD can win you good salaries.
• An industrial PhD. Salaries are good with industrial PhD. However, in my opinion there is no major difference between doing a project for the company and doing an industry funded PhD. The work tends to be too applied with the funding company often expecting results that can benefit the company or its products.
• Hot topic. Just because deep learning is a hot topic, it may be extremely tempting to start a PhD in it. By the time you finish, the academic community might have moved on and be churning out papers on a different area of your field. So, sticking with something exciting for you is way better to survive the PhD grind.

So, simply choose a topic that interests you and will keep you going for years. It is quite normal to drift away from where you started. But even with the drift, I have always seen students stay in the expert zone of their supervisor and not drift too far.

Location

By location I mean the town / city where you will be working. Most of my friends declined PhD positions simply because the lab is located in a very small town in some corner of the world.

• It is worth noting that some extremely skilled scientists and professors prefer to work in such quiet towns for personal reasons. I tend to think that taking a quiet stroll along the streets and parks of such small towns can trigger creative ideas.
• For a student, one advantage of working in such locations is that the cost is less. Given the salaries that PhDs get, you get to save a significant sum even during your PhD by working in such locations.
• There is no chance to party every Friday night. So you get to focus more on your thesis which by itself is challenging enough to finish in 3 years. With partying and extra curricular activities you are only extending the life of your PhD.

So why bother much about where you work when you get a chance to work with a renowned supervisor on a highly interesting topic. If you do best during your PhD years, you can bag a post-doc position in the city of your choice 😃.

Size matters

Some research groups tend to publish on diverse topics. For instance a group may be publishing in Computer Vision, Robotics and Medical Imaging. Such a diverse group indicates either one of the two:

• Your supervisor is a world renowned Professor who can manage such a big, diverse group by hiring very skilled researchers.
• The group is inter-disciplinary and so they need to work on multiple fields.

A bitter fact in such a big group is that the supervisor simply cannot keep up with what is happening with every student. For instance it becomes difficult for a Professor to read papers in every single conference in CV, ML, NLP. So they tend to manage the group as a hierarchy. They hire post-docs who are experts in different areas. The PhD students in turn get guidance from the post-docs. It is very easy to identify such groups looking at the publication record of your potential supervisor in the past 3–5 years.

• If the group is hierarchical, talk to the post-doc who is going to be your supervisor. Because you will mostly discuss your problems with your post-doc. Nevertheless, be prepared to work all by yourself in this kind of environment as your post-doc will be busy in meeting, writing grants, etc.
• If the group is small and focused on 2–3 areas of a field, then you can expect much more help from your supervisor. He/she will be keen to get good publications out of you which in turn bring grants, which in turn expand the group and the cycle continues…

“Small is beautiful” when it comes to group size. But if you can do it all alone, then choose a large group, work independently with lot of independent work happening around you.

Publishing style

In my opinion, there are two kinds of publishers in academia: quality and quantity publishers.

• Quantity publishers wish to publish a lot of papers in a given year. So they send their papers to workshops or a tiny unknown conference in some corner of the world. To them what matters is getting another paper in their list of publications.
• Quality publishers tend to focus on the quality of papers. So they wait and send the papers to top notch conferences or journals. For this, they patiently do lots of experiments to convincingly argue the statement they are making in the paper. Needless to say, they are the ones that get the most citations as they are constantly pushing the state-of-the-art. So their work tends to be recognised internationally.

So, if you find your potential supervisor has 30 publications in a year, be sceptical. A supervisor with 3–5 publications in a Tier 1 conference or journal could be a better bet to work with simply because you will also be expected to publish quality papers and not churn-out quantities.

Freedom

One of the joys of doing a PhD is that as a creative individual, most of the times you get to do whatever you want, at least within the scope of the area you are working on. But it may not be this way:

• Your supervisor may be strong about his own ideas and could always expect you to try and experiment with his ideas.
• Or he/she can be someone who lets you lead ‘your’ work and just guide you whenever you are totally lost or need resources like more compute power or some equipment to run your experiments.

First, think twice what sort of person you are. Are you abreast with what is happening in the field? Can you come up with your own ideas and show through your experiments that the ideas are worth it? if so, you probably need to work with the latter style of supervision. It is worth talking to your potential supervisor’s team members to find out how they come up with ideas, implement them, publish and iterate the process.

Group Mates

Though a PhD is a solo journey with mounting pressure to publish as days pass by, you tend to spend a lot of time with your group mates, go for a few drinks with them every now and then. They may not be too helpful to you even if they want to as their expertise will be in a different area. But they do define the energy and morale of the team. The team could bring positive vibes to you by simply encouraging you in your journey. Or they can simply throw words at you during that 5 mins lunch break chats which can tumble the bricks of your confidence. So the energy and morale of your potential group mates is never to be underestimated while choosing to commit yourselves 3+ years.

Try to chat with the potential group mates as much as possible before choosing to commit working with them for few years. Identify the thin line of work culture running in the group.

Other Commitments

Some other commitments for a PhD student could be:

• Teaching courses and preparing course materials
• Supervising MSc thesis
• Supervising students in labs

All the above skills are quite valuable for someone wishing to pursue an academic career. But they also consume valuable research time from your 3 years. Would you like to graduate as that all-rounded person who has tasted every aspect of academia? Or would you like to focus on pure research during your PhD and let alone teaching and supervising others for later stages of your career?

Expectation

With all other factors fixed, some supervisors simply expect the students to work all by themselves, come up with ideas and get results. This is because a PhD is the only time to mould you into that independent scientist who can constantly push the field ahead.

Some supervisors will help you quite a lot. They discuss with you all the ideas you need to experiment with. They write the paper if you are running experiments. The do the figures for your paper if you are coding the solution.

You and “only you” know who you want to work with. You may asses yourself by looking at how you behaved while doing your MSc coursework. Did you work alone? Did you like group projects or independent projects?

Conclusion

Unlike other degrees, the value of a PhD is in what you contribute! You are more likely to succeed if you choose to work on the right topic and the right people. All else is next. I hope this article helps make that right decision before embarking on that journey rather than regretting with a hindsight. Happy PhD! 😄

Also note: Publishing a few papers as the first author is the starting point for a research career. PhD is one defined path to do it. But in a small world, don’t many paths lead to the same destination? 😏