Doing PhD at the Center for Machine Perception

Tomáš Pajdla, Václav Hlaváč, Jiří (George) Matas

http://cmp.felk.cvut.cz

Der Mensch kann nur Mensch werden durch Erziehung.
Er ist nichts, als was die Erziehung aus ihm macht.
I. Kant

• What is PhD.
  During the PhD studies at CMP, PhD students (i) learn how to do scientific work, (ii) solve a new research problem, and (iii) adopt culture of the scientific community. The studies are finished by submitting a PhD thesis which describes the student's contribution in his/her field of expertise.
• Benefits of PhD students.
  
  1. PhD students at CMP work in an open environment where a discussion and an initiative are welcome.
  2. Students look for and solve their research problems in a cooperation with their supervisors and colleagues from CMP and elsewhere.
  3. Students have opportunities and are encouraged to take part in research exchange programs with  cooperating foreign laboratories.
  4. Students are supported by infrastructure and funding from CMP to do their work and present it at prestigious conferences around the world.
  5. Stipends are supported by CMP financial resources. The stipend paid depends on the student's performance and can raise up to the legal maximum.
• Obligations of PhD students.
  PhD students are required to:
  1. (applies to full time PhD students) work full time on their PhD thesis. Students are required to consult their potential extra jobs with their supervisor and ask for approval. Students working on a defined research project are obliged to cooperate with the project leader in order to achieve projects milestones. Students not working on a well defined project help other running projects according to their abilities if needed.
  2. inform all the members at CMP about the progress of their work once a month in the form of a ``PhD report'' as described at the report instruction page.
  3. publish their results in the form of (i) presentation at the internal CMP seminars, (ii) research reports, and (iii) conference and journal papers.
  4. teach at CMP as they are obliged by PhD bylaws at CTU, Prague.
  5. improve their professional abilities by attending CMP seminars and some lectures given at CMP.

• Finishing in 3 years: recommended schedule and related milestones.
  The stipend of a full time PhD student is available for three years only. Thus the thesis should be finished in three years. Meeting the deadline is not easy. A prototype schedule dividing the PhD study into three year-long periods is presented to help the PhD candidate and his advisor to plan the study and check how they perform in its run.

  Y	M	Activity	Milestone
  1		Looking for the topic
  	1		Supervisor proposes area of research
  		Attend the ``Doing PhD at the CMP'' lecture
  		Learning CMP Environment
  		Reading and annotating skills	Annotated bibliography created, maintained throughout the PhD
  		Presentation skills
  	4		Internal seminar: the most relevant work by others
  		Writing the state of the art report
  	8		The state of the art written
  	9		Internal seminar: the main contribution proposed
  		Reviewing the main contribution
  		Writing the CVWW paper
  	12		CVWW paper submitted
  2		Working on the main contribution, 'core of the thesis'
  	1	Writing the first conference paper	Presentation at CVWW
  	2		The first conference paper submitted
  	3	Writing the thesis proposal report	The thesis proposal report written
  	6		The thesis proposal defended (Minimum)
  	7		Presentation at the first conference
  	11		Internal seminar: defense of key results
  	12	Writing the second conference paper	Rigorosum examination passed
  3		Writing the thesis
  	2		The second conference paper submitted
  		Writing a journal paper	Outline, SoA, Goals written
  	5		The journal paper submitted
  	6		Theory, Methods written
  	9		Experiments written
  	11		Abstract, Introduction, Conclusion written
  	12	External seminar: the results of the thesis	The thesis submitted
  		Internal seminar: thesis defense training

• How to do research for PhD
  To do science and research well is very difficult. A good definition of science and excellent examples of correct approaches to the scientific work can be found in works of Karl Popper [6,5] and Richard Feynman [3]. Please, read some and better all of them!

  The principles of sciences differ from discipline to discipline but the essence of all sciences is understanding. Different sciences use the understanding for different purposes. Natural sciences to predict future observations, technical sciences to solve problems, historical sciences to explain the past, economical sciences to steer the economy, political sciences to control the society, and philosophy to draw most general principles.

  The key goal is therefore to understand more. How to do it? It is exactly what students learn during their PhD studies.

  • CMP seminars
    CMP organizes internal and external seminars. Internal seminars are informal meetings where ideas and preliminary results can be presented, discussed, and criticized before they are published. Seminars also serve to disseminate knowledge among CMP members and to master the presentation of results. External seminars provide the opportunity to present already finished work to a larger community. It also gives a forum for CMP visitors to present their work.
  • Publications
    Publications are the ``materialized output'' of most scientific work. The unwritten law ``publish or perish'' makes scientists publish in order to justify financing their work. However, though publishing is important, only good work deserves publishing. For a PhD student, it is better not to publish at all than to publish at all costs or to publish incorrect, incomplete, or previously published work.

    Scientific publication must contain a new knowledge. As a definition of a novelty the following citation from the instructions for referees of IEEE Transactions on PAMI can be adopted: ``The novelty will usually lie in new results, methods, observations, concepts, or applications, but may also reside in syntheses of or insights into previously reported research.''

    Writing publications right is difficult. Read [4,2] to find more about how to write and publish a scientific paper. Regarding publications at CMP, they have to follow CMP publication rules. Please, never forget to acknowledge research grants. Each publication has to be approved by student's supervisor.

  • Links to good practice advices related to PhD study.
    It is very reasonable to learn from experiences of successful PhD students of the best universities world wide. We point you to two papers written by students. The first one, entitled ``How to do Research At the MIT AI Lab'' was edited by David Chapman in 1988. The second one ``How to Be a Good Graduate Student'' was written by Marie desJardins in 1994.

    There is a chapter available from the Richard M. Reis' book ``Tomorrow's Professor'' entitled ``Your Professional Preparation Strategy'' that is worth reading too

    The University of Queensland, Australia, has collected advices for PhD students at their university professionally.

    An approach to good supervisory practice is described in [1].

• PhD students graduated from CMP.

Literatura

1

Dorman Sir et al. Christopherson.
Research student and supervisor, an approach to good supervisory practice, May 1997.
CMP.techrep.T178.

2

Fobert A. Day.
How to Write Publish a Scientific Paper.
Oryx Press, 3rd edition, 1988.
CMP.book.A127.

3

Richard Phillips Feynman.
The Meaning of It All: Thoughts of a Citisen Scientist.
Addison-Wesley, Reading, Massachusetts, 1998.

4

Herbert B. Michaelson, editor.
How to Write and Publish Engineering Papers and Reports.
Professionals Writing Series. Isi Press, Philadelphia, USA, 2nd edition, 1986.
CMP.book.YC31.

5

Karl Raimund Popper.
Alles Leben ist Problemlösen.
Piper, München, 1994.
in czech: Zivot je rešením problémøu, Mladá fronta 1998.

6

Karl Raimund Popper.
The Logic of Scientific Discovery.
Routledge, London and New York, 1995.
in czech: Logika vedeckého bádání, OIKOYMENH, 1997, CMP.book.D3.