How to find your advisor
I had tweeted this earlier about "Rocking your PhD":
It is that simple. This is actually a concise version of a longer advice I provided earlier.
Since I haven't talked about it before, I like to now write some suggestions on finding an advisor.
In the first semester of your graduate studies, take 3-4 classes you are interested in. This provides a good opportunity to meet and impress your prospective advisor. If there is a class project, go overboard and exceed expectations. Try to improve on an algorithm mentioned in the class, and discuss this with the prospective advisor.
Before you commit with an advisor, make sure it is a good match. As you take a class from your advisor, see if you can work together during the semester on a small entry-level problem/project. Because after you commit working with an advisor, it is messy/costly to switch advisors, since there is a sunken cost of investment from both parties.
You want to find a hardworking advisor, so you can work hard as well. Assistant professors have a lot of incentive to work hard to get tenure. On the other hand, many Associate and Full Professors also work hard, and may provide additional benefits: more experience and an established research group where you can get mentoring from senior students. In any case, you should check the track record of the faculty, giving emphasis on recent publications.
Try to find an advisor that you can learn a good toolkit, e.g., formal methods, graph theory, probability theory, or machine learning theory. Application domains come and go following trend cycles, on the other hand, toolkits are fundamental and can be applied to application domains as needed. To give a personal example, after learning formal methods and reasoning about distributed algorithms as my toolkit, I was able to apply them to wireless sensor networks domain and later to cloud computing domain, distributed databases domain, and hopefully soon to the distributed machine learning domain.
To balance the previous advise, make sure the advisor is also doing work on the practical side of things. Faculty jobs are very sparse, and after your PhD, you should have the option of joining industrial labs/positions.
I think it is necessary but not sufficient.
Can you succeed with a bad advisor?
First, define "succeed". If it is a good faculty job or a prominent industrial research position, I don't know examples of this. It may be possible, if there is good support from other graduate students/faculty.
Are there bad advisors?
There are definitely advisors that don't care. Or that are too busy and don't help enough.
Is PhD largely an apprenticeship?
Yes, I think so. You learn by diffusion your advisor's taste in finding problems and devising solutions. In CS there are distinct categories of theory, systems, metrics, engineering, and algorithms people. I think the advisor imprints on the student not just on the research category but also perspective on things, such as openness to new things, having a home conference versus being more promiscuous, etc.
Is there a personality match thing?
Even though this is a professional relationship, personality clashes between the advisor and student may impede progress, or sometimes even lead to a blow up. A student that is very stubborn/uncoachable is not good, but the student shouldn't just follow or wait for instructions either. I like the student to push back and defend what he/she thinks is true, but also to be open-minded/receptive to alternative opinions/perspectives.
I knew of a faculty who asked students for a Myer-Briggs test (the one where we all get INTJ category). He was a very smart professor, so probably he found utility in that. There is also the big-five test. Personality styles may be useful to gauge how the student can fare with different advisement styles: socratic method, sink or swim method, trainer method, coaching method, and manager method. As for me, I don't care to know personality types of my PhD students, I just want to see if we can work well, discuss well, be productive and grow together.
- Find a hardworking advisor
- Get a senior PhD student mentor
- Read a lot of papers critically
- Write a lot, get feedback
- Publish your first paper early to build confidence
- Publish your 2nd, 3rd, 4th papers
- To sustain, exercise regularly
It is that simple. This is actually a concise version of a longer advice I provided earlier.
Since I haven't talked about it before, I like to now write some suggestions on finding an advisor.
How to find your advisor
Ask around and get advice from senior PhD students in the department about faculty as potential advisors.In the first semester of your graduate studies, take 3-4 classes you are interested in. This provides a good opportunity to meet and impress your prospective advisor. If there is a class project, go overboard and exceed expectations. Try to improve on an algorithm mentioned in the class, and discuss this with the prospective advisor.
Before you commit with an advisor, make sure it is a good match. As you take a class from your advisor, see if you can work together during the semester on a small entry-level problem/project. Because after you commit working with an advisor, it is messy/costly to switch advisors, since there is a sunken cost of investment from both parties.
You want to find a hardworking advisor, so you can work hard as well. Assistant professors have a lot of incentive to work hard to get tenure. On the other hand, many Associate and Full Professors also work hard, and may provide additional benefits: more experience and an established research group where you can get mentoring from senior students. In any case, you should check the track record of the faculty, giving emphasis on recent publications.
Try to find an advisor that you can learn a good toolkit, e.g., formal methods, graph theory, probability theory, or machine learning theory. Application domains come and go following trend cycles, on the other hand, toolkits are fundamental and can be applied to application domains as needed. To give a personal example, after learning formal methods and reasoning about distributed algorithms as my toolkit, I was able to apply them to wireless sensor networks domain and later to cloud computing domain, distributed databases domain, and hopefully soon to the distributed machine learning domain.
To balance the previous advise, make sure the advisor is also doing work on the practical side of things. Faculty jobs are very sparse, and after your PhD, you should have the option of joining industrial labs/positions.
MAD questions
Is a good advisor enough?I think it is necessary but not sufficient.
Can you succeed with a bad advisor?
First, define "succeed". If it is a good faculty job or a prominent industrial research position, I don't know examples of this. It may be possible, if there is good support from other graduate students/faculty.
Are there bad advisors?
There are definitely advisors that don't care. Or that are too busy and don't help enough.
Is PhD largely an apprenticeship?
Yes, I think so. You learn by diffusion your advisor's taste in finding problems and devising solutions. In CS there are distinct categories of theory, systems, metrics, engineering, and algorithms people. I think the advisor imprints on the student not just on the research category but also perspective on things, such as openness to new things, having a home conference versus being more promiscuous, etc.
Is there a personality match thing?
Even though this is a professional relationship, personality clashes between the advisor and student may impede progress, or sometimes even lead to a blow up. A student that is very stubborn/uncoachable is not good, but the student shouldn't just follow or wait for instructions either. I like the student to push back and defend what he/she thinks is true, but also to be open-minded/receptive to alternative opinions/perspectives.
I knew of a faculty who asked students for a Myer-Briggs test (the one where we all get INTJ category). He was a very smart professor, so probably he found utility in that. There is also the big-five test. Personality styles may be useful to gauge how the student can fare with different advisement styles: socratic method, sink or swim method, trainer method, coaching method, and manager method. As for me, I don't care to know personality types of my PhD students, I just want to see if we can work well, discuss well, be productive and grow together.
Comments