The Supervisor Role

I will try to sketch here how a supervisor may behave toward his students. The article will have naive thoughts initially. However, as I proceed more in my PhD, I expect by 1 Jan 2019 the article should have more strong and practical thoughts.

Before Each Semester

Before student arrival (first semester) or any semester, there should be communication to inform the student about the expectation of the next semester (or better student communicate first). Sometimes there are funds and student can start directly in research. In other times student has to work as TA to support himself financially. Usually PhD student has some courses and the best for him to finish these courses as soon as possible.

It is important that supervisor don't abuse the student availability. We may assume 40 hours per week distributed on selected activities fairly. Supervisor may ask the student to sketch a schedule for the approximate hours coverage per week.

Courses: Supervisor should help the student in picking the suitable courses for him. Guiding the student to make course projects in similar things to his research line (student may talk to lecturer about that).

The First Meeting

Supervisor should meet the student as soon as possible. Potential things:

  • What is the fund status?
  • What are the current lab projects...How they are connected.
  • Listen to the student to know what is best for him..understand more his interests.
    • Student should understand that sometimes the Fund from some destinations means doing some work for them.
  • Asking the student to talk with every member in the lab about his work / papers so far to build vision.
  • Communicate with him again to determine what should be the first thing to start with it.
    • Student should know soon he may derive the remaining for the research road...not to wait decisions and inputs always.
  • Establish a repetitive meeting to get student status and help/guide him. May be also he asks for progress reports regularly.

Roles: Resolve conflicts and problems

  • From time to time a student may face problems, the supervisor should help in resolving them.
    • May be financial problems
    • May be technical problems
    • May be he doesn't like current research direction
  • May be also student causes problems and supervisor may need to solve that wisely
    • E.g. A girl in the lab felt embarrassed when her college refused to shake hands
    • Supervisor may ask to do so in gentle way
    • May be ask to communicate reasons
    • May ask him if he has further concerns

Roles

Team Communication

ToDo

Papers Reading

ToDo

Directing Student Researh

  • This is the most critical part of supervisor role.
  • Supervisors are of different types
    • Those who doesn’t exist. No direction. Student depressed usually and paying costs of his wrong research decisions.
    • Those who exist in the picture, have shallow knowledge and their help usually meaningless or destructive
    • Those who are doing the right job toward their students.
  • Assume, PhD student has shallow knowledge about the field, or even specific problem.
    • Think for a while if student found idea that already implemented. He doesn't know all literature.
    • He spent a year and submitted work to conference and received rejection as most of this work already there..no novelty
    • A good supervisor will save student from such wrong path
  • Assume, PhD student picked an interesting problem to work in it.
    • However, some giants (e.g. Google and Facebook) detecting much of resources for the problems
    • They already have great results and still fighting for more
    • Student worked hard for 1 year and ended with good results...but far from what the giants achieved!
    • A good supervisor will save student from such wrong path. "Don't compete with Giants Directly"
  • Assume, PhD student working in an interesting problem, thinking in some ideas and found some of them.
    • However, The testing datasets are of size Tera Bytes. We need some clusters. Machines with strong GPUs!
    • Students work around to test his code ended at nothing! (e.g. trying to use CPU to test in 4 weeks instead GPU in 4 days)
    • A good supervisor understand nature of problems and the needed computational resources for it. If we can't do it, let's try problems where we have available resources
  • Supervisor should help student in picking the right problem WHERE he has some experience in the problem.
  • If no data are available for a problem, should guide students while preparing (pre-processing) data. If we have wrong data, we will have wrong results!
  • Supervisor should point to the critical skills needed for success.
    • E.g. importance of reading many papers. However, learn to read faster. Skim papers in 10 minute and see if related or useful.
    • Going to present work in conference? What to do? What not to do?

Literature Awareness

  • For a proper research directing for students, advisor must be aware with the literature.
  • He also must be aware with the new potential research direction
    • Imagine a supervisor is still working in features extraction while literature is consistently moving toward learning features by deep learning.
    • A supervisor who notice the literature important changes in the early stages, could lead the community in this new direction.

Getting Funds / Resources

  • One of supervisor roles is the availability of needed resources. For example, we may need in labs more storage or new strong GPUs for our processings.
    • Sometimes could apply for grants (e.g. NVIDIA GPU Grant).
  • Write proposals for different destinations about our future research