Graduate Student FAQ

QC1. Do you have money for new RAs?
QC2. May I do an independent study project with you?
QC3. Will you be my (non-thesis) advisor?
QC4. What courses should I take?
QC5. Will you be on my committee?
QC6. If you're on my committee, how involved do you want to be, and when should I contact you?
QC7. As my committee member, what instructions do you have on the preparation of my thesis, and preparation for my thesis defense?
QC8. What are your research areas?
QC9. Will you be my MS Thesis advisor?
QC10. I'm a current PhD student looking for an advisor. May I sit in on your PhD group meetings to see who your students are and what their topics are?
QC11. Will you be my PhD advisor?
QC12. How will you help me if you become my advisor?
QC13. How do I know if I'm suited for research?
QC14. How do I get started on research (i.e., on what should I work)?
QC15. What if I have my own ideas about what to work on for a thesis topic?
QC16. Where can I find information about the requirements for the degree, and about requirements / standards for the thesis?
QC17. What are your requirements for a thesis?
QC18. How long does it take to get a MS with thesis, or a PhD?
QC19. Can I do a co-op or internship even if I'm going to start, or already working on, a thesis?
QC20. How do I become familiar with what you do?
QC21. What do I need to know to work with you?
QC22. What software will I need to use?

QP1. What are my chances of admission, can the fee be waived, do I really need GRE scores, ...?
QP2. Will you recommend or secure my admission?
QP3. Will you support me?
QP4. How do I get support?
QP5. Will you be my advisor (i.e., may I join your research group)?
QP6. What should I do to get off to a good start at NC State?

QA1. How should I contact you?
QA2. How should I send you papers, information, etc.?

QC1. Do you have money for new RAs?

A: No. I am already working with students who need support.

QC2: May I do an independent study project with you?

A: Maybe. Good independent study projects are harder than regular courses, require a serious commitment of effort, and cannot be left as the effort that gets cut when regular courses become demanding. Most students: (a) seriously underestimate how difficult research is, (b) seriously overestimate how prepared they are to do research, and (c) do not budget enough time to do something worthwhile, original, and of good quality. I'm willing to talk about such projects if you understand the work involved. Plan on having to complete a substantial, well-written paper on your study in order to receive a passing grade.

QC3: Will you be my (non-thesis) advisor?

A: Yes. Non-thesis students don't need an advisor, they just need a few questions answered. I will answer questions during office hours, whoever you are. Please don't bombard me with emails.

QC4: What courses should I take?

A: You'll have to meet me during office hours and tell me your goals and your background for me to help you.

QC5: Will you be on my committee?

A: Maybe. First you should discuss with your thesis advisor the choice of committee members. If your advisor suggests or agrees to my name, then the answer will likely be "yes." Occasionally the answer will be "no" because a) I'm swamped, and/or b) I'm not really a good choice to understand your topic. Please don't take it personally if I need to turn you down.

QC6: If you're on my committee, how involved do you want to be, and when should I contact you?

A: I'd like to hear about your thesis project at the time I agree to be on your committee, but typically after that would see you next only at the defense. In some cases, a committee member is actually helpful to the student in his/her work, in which case they meet as needed. "Pinging" your committee member, in person or email, every 6 months just to let them know you are still alive and making progress is a good idea.

QC7: As my committee member, what instructions do you have on the preparation of my thesis, and preparation for my thesis defense?

A: If I am on your committee (but not your advisor), and you have an exam coming up in which I will participate, please help me do my job by following these guidelines, in roughly the order given:

1. Read and follow instructions posted by the graduate school and your department for theses, scheduling exams, etc. Be sure you are aware of graduate school deadlines, which are long before the end of the semester; also allow for the lead time the graduate school requires for scheduling exams (PhD: 3 weeks!). Some online resources are:
   • The Graduate Administrative Handbook
   • Graduate Student Resources
   • Thesis and Dissertation Information
   • Computer Science Handbook for Graduate Students
2. Get your thesis or proposal in good shape and approved by your advisor before giving it to me. The advisor is the first (and most important) person to give you feedback, and to satisfy.
3. Contact me (and other committee members) with 2 or 3 dates you prefer, and I'll tell you my available times on those dates. When you find a compatible time for all committee members, notify me immediately. There is often a flurry of students trying to hold exams at approximately the same time (right before the grad school deadline). First one to nail down a date and a time wins!
4. Give me a paper copy of your completed thesis or proposal one week before the exam date. Please also send an electronic copy by email. I will read the paper copy. I'll use the electronic copy only if needing to look at it when I don't have the paper one readily accessible.
5. The thesis or proposal should already have been proofread by a careful, qualified reader for spelling, punctuation, and grammar errors before I see it. Occasional minor errors are no problem, but if the thesis or proposal has serious readability problems I will return it for revisions. I will not be able to understand and appreciate the content if the writing is seriously deficient.
6. After turning over your thesis or proposal to your committee, but before your actual exam, please review your slides with your advisor. With no interruptions or questions, your presentation should be approximately 30 minutes (MS) or 60 minutes (PhD) long; the actual exam will be longer due to questions and discussion. This equates to a presentation with roughly 20 (MS) to 40 (PhD) slides. Long presentations are invariably cut short on the spot by the committee, which is flustering, but entirely avoidable.
7. It is often helpful to schedule a brief meeting one-on-one with your committee members before a defense, to orient your committee members, and/or to preview issues that may be raised in the exam. This step is optional; some do it, some don't.
8. The exam is a formal event. It is the policy of the College of Engineering (with which I agree) that graduate students may not provide refreshments to committee members. Please do not plan on food at your exam; that is not the purpose, and it strains the independence between student and examiner that ensures you are treated professionally. If the advisor wants to provide food, that is OK, although my belief is that the time for refreshments is after the exam, not before it.
9. You should plan on some ``revision time'' following an exam. It is possible you will need to make some changes to your thesis, based on feedback from your committee. It is risky to expect no changes, and to schedule something major (like leaving town for a new job) immediately following an exam. In most cases, such changes require only 2 or 3 days to implement and get approved by the committee.

QC8: What are your research areas?

A: My primary research interests center on finding mathematical formalization of agents, knowledge, reasoning, learning, and decision making that increase understanding of these topics and that support effective application of ideas to practice. I am currently working on problems of:

1. Preference representation, reasoning, acquisition and adaptation for qualitative decision theory;
2. Mathematical foundations of rationality and reasoning;
3. Mechanical and physical foundations of computation and psychology;
4. Knowledge representation and architectures for rational, intentional, and introspective agents;
5. Foundations of informational, personal, and national security; and
6. Knowledge-based monitoring and alerting.

I am also becoming interested in areas including

1. Rational data exploration and visualization
2. Foundations of bioinformatics

QC9. Will you be my MS Thesis advisor?

A: Maybe. Become familiar with my work and then talk with me during office hours. I may make suggestions about other people to talk to, or if there is substantial promise of a mutually interesting topic, suggest we continue to talk. Ordinarily I will require a signable thesis proposal before agreeing to become a thesis advisor.

QC10: I'm a current PhD student looking for an advisor. May I sit in on your PhD group meetings to see who your students are and what their topics are?

A: Currently I do not have regular group meetings.

QC11: Will you be my PhD Thesis advisor?

A: Maybe. Become familiar with my work and then talk with me during office hours. I may make suggestions about other people to talk to, or if there is substantial promise of a mutually interesting topic, suggest we continue to talk. Ordinarily I will require a signable thesis proposal before agreeing to become a thesis advisor.

QC12: How will you help me if you become my advisor?

A: I expect to give students a lot of freedom to define what they work on and how they proceed, but at the same time to redirect them if they are not making progress. I may provide equipment, space, and some funding for travel, etc. I encourage meetings between students and collaborators. At the same time, I will not pick the problem for a student; that constitutes part of the process of research.

QC13: How do I know if I am suited for research?

A: If you are contemplating a thesis and have never looked at research papers, who knows? If you read the research literature (conference proceedings and journals) and get excited about the work described, that is a good sign. It's like cooking: if you love eating good food, you may be a great cook, but if you don't like good food, why be a chef?

At the same time, getting excited by ideas does not always guarantee success. Research requires hard sustained effort. To some extent doing research becomes easier the longer you do it, but at the start it requires effort and uncertainty unlike any most students face in regular coursework. No one gives you a problem, or if they do, you cannot assume it has a solution in the form given. You will get discouraged at times during a graduate career. Find a problem interesting enough to keep you motivated during these hard times. Make it your own problem; doing something because someone else finds it interesting risks working hard only to give up some night when you realize the problem bewildering you is someone else' problem, not one that means something to you.

QC14: How do I get started on research (i.e., on what should I work)?

A: There are many ways of proceeding, depending on how much you have already read and done. Here are some general suggestions.

1. Having an idea of your interests, identify the high-quality conference or workshop or journals that more or less specialize in that topic. At present, artificial intelligence spans many areas, with many journals, conferences, and workshops; see the AAAI AI resources page for pointers. To find the papers to study, try the following:
   • Copies of journal issues and proceedings are in the library. Top journals include
     • Artificial Intelligence
     • Computational Intelligence
     • Journal of AI Research (JAIR) (online, free)
     • Journal of Machine Learning Research (online, free)
     Top conferences include
     • National Conferences on Artificial Intelligence (AAAI)
     • International Joint Conference on Artificial Intelligence (IJCAI)
     • Knowledge Representation and Reasoning (KR)
     • Uncertainty in Artificial Intelligence (UAI)
     • International Conference on AI Planning Systems (AIPS)
     • Theoretical Aspects of Rationality and Knowledge (TARK)
     • Knowledge Discovery in Databases (KDD)
   • All IEEE and ACM papers are available online to NCState faculty and students. See http://www.lib.ncsu.edu/eresources/computer-science.html for access to these and other databases.
   • Programs (and often abstracts) of recent or upcoming issues and proceedings are available online (e.g., search on "KDD 2002" to see the most recent program). The actual papers, even if not available from the society yet, can often be found on the author's website by searching on the paper title.
   • Citeseer (http://citeseer.nj.nec.com/cs) is an invaluable link to research in computer science and engineering.
2. Peruse the last year or two of the proceedings or journal. This gives you a sense of what the current research topics are, what the required background is, and what the standards for a research contribution are. It also tells you whether you like the topic, have an aptitude for it, are qualified to work in it, and are motivated enough to seek new, creative solutions.
   To "peruse" does not mean to read every word of a 1500 page conference proceedings. Rather, you are skimming the table of contents to see what the scope of papers is, and what sessions or topics might be of interest. Then, you are restricting your attention to those papers. You read the title and abstract, possibly skim the figures, and the conclusions and possibly references. Now you know something about the problem, the solution, and how it was accomplished. If you do this for a conference, or for a year's worth of a journal, you get a feeling for what is going on in research, without being overwhelmed.
3. Pick two or three papers that you particularly liked and are interested in. Read them in depth and understand them. Then bring those papers, and your opinions and ideas about them, to discuss with me. I will ask you questions, find out what your motivation is, make suggestions about alternatives you may not have considered, and in general explore the topic with you. Even better is for you to write an analysis of the papers for me or fellow students to read.
4. Although you should become familiar with my own research interests and might well read some of my own publications if you want to work with me, under no circumstances should you expect to read only my publications in finding a research topic. If you do not attempt to understand what I do, one might conclude you are not serious about working with me. That might not be a problem, as I might still be able to help in identifying a faculty member who might be an appropriate advisor. However, if you do not attempt to understand what others have done, one might conclude you are not serious about research. That certainly will be a problem, at least if you expect to complete a degree.

Following something like the procedure just described will save a lot of time all around. Regardless of whether it leads to a thesis, you will benefit from the time spent; for many students, it's the first encounter with the research literature they have ever had.

You may think "but what what are you doing to help me?", and the answer is that I guide you, not push you from the rear. Identifying and defining the research problem is an essential part of the graduate experience, particularly at the PhD level. Indeed, once you do find a good topic, it is up to you to work hard on mastering the relevant background and literature. I can offer suggestions and help you interpret what you find, but the burden is on you to learn and exploit the material, not on me to teach it to you. The same goes when solving your problem, and when writing a thesis. I can offer suggestions and critique drafts, but I will not solve your problem or write your thesis for you.

The biggest problem facing many students is understanding that graduate research is completely different than undergraduate studies. As an undergraduate, one expects to be assigned problems that have solutions, that require known amounts of effort to solve, and that will be explained in lecture. As a graduate student, one typically is not assigned a problem, but must find it. One does not know how much effort will be required to solve the problem. The problem will not be explained in any lecture; instead you will have to educate yourself in the literature and needed background. Indeed, your thesis defense will be the first lecture on the problem and its solution. The student who enters graduate school expecting to be taught everything he or she needs to know puts himself or herself at a disadvantage.

As a graduate student, one should always stay occupied with trying to understand and solve problems. One cannot choose a thesis topic if one does not understand what are the real problems, as opposed to the superficial problems. One cannot choose a thesis topic if one cannot tell how interesting are the problems, and how much aptitude one has for them, or what background one needs. Working on a problem or portion of a problem reveals many of these things in a way that helps one choose how to proceed. To complete a thesis project, one needs this knowledge. One must therefore work at it all the time. People have been known to be struck with inspiration that identifies an important problem and its solution; but such inspiration comes most frequently to those who have prepared to receive it with persistent self-motivated study.

The ability to read and think critically forms a key element of success as a researcher. In particular, when reading a paper or part of the literature, the student needs to keep asking the following questions.

• What is the problem being solved?
• What information or conditions are assumed?
• What does this information mean?
• Are the assumptions reasonable? Consistent? Complete?
• Are there other reasonable assumptions?
• Does the solution extend to work under different assumptions?
• What qualities or characteristics are desired?
• Are the desired characteristics reasonable?
• Are there other reasonable desired characteristics?
• Does the proposed solution actually solve the stated problem?
• Does the proposed solution solve other problems?
• What problems are left unsolved?
• What problems are left unstated?
• Can I formulate the problem in precise mathematical terms? Some portions? If not all, why not?

The student needs to ask these questions about every paper of interest, including those I have written, or risk failing to understand the material

Never stop reading, and never assume you have read enough. If I suggest looking at a paper, do not read it and stop. Instead, use your understanding of it to help you identify the next paper or papers to read, whether from among the papers cited by the article or by searching for uncited papers on related topics. If you ever think you have read everything, I will be happy to tell you something to read next.

QC15: What if I have my own ideas about what to work on for a thesis topic?

A: Wonderful! Be ready to discuss and justify them, and possibly to refine the ideas as a result. Even if your ideas don't work out, it shows me you have initiative and enthusiasm.

Note: a follow-on question might be "If I bring my own ideas, does that mean you won't consider me for funding"? If you work on something of interest to me, you can consider it to be "fundable".

QC16: Where can I find information about the requirements for the degree, and about requirements / standards for the thesis?

A: See question QC7 above.

QC17: What are your requirements for a thesis?

A: For an MS thesis, I expect 1 or 2 publications and possibly data and software. For a Ph.D., I expect several publications and to gain new insights, and new opportunities.

Published work is essential. I will not advise theses whose only significant output is "I learned a lot"; that outcome characterizes independent study courses, not theses. Expect to submit papers before you graduate. It helps to keep in mind the submission deadlines for conferences in the area. Often a conference paper provides the blueprint for a thesis.

QC18: How long does it take to get a MS with thesis? to get a PhD?

A: There is no formula for degrees, and times can vary greatly. A "typical" MS student starts in fall of year Y, takes two semesters of course work and begins working on a thesis in summer of Y+1, then completes two more courses and the thesis by the end of spring Y+2, or (frequently) early summer Y+2. Think 2 years or a bit less.

A "typical" PhD student (already having a MS), if there is such, takes courses for 1 year, starts research and finishes any remaining courses in the third semester, and works full time on research for another 2 years. Now we're at 3.5 years. Five years is not unusual. Theoretical (mathematical) theses can take less, but it takes an unusually good student and fortuitous choice of problem to achieve that.

QC19: Can I do a co-op or internship even if I'm working on or going to start a thesis?

A: Early in your research, yes. Having more experience, particularly in a research lab, is a plus both before and after graduation. Later in your research is probably not a good idea; you lose momentum and delay graduation (perhaps forever) in this case.

QC20: How do I become familiar with what you do?

A: Read some of my papers and the answer to QC8 above.

QC21: What do I need to know to work with you?

A: It depends. Some problems require only basic knowledge of logic, discrete mathematics, and standard computing topics. In general, though, AI problems involve questions also studied in detail in advanced logic, philosophy, mathematics, economics, linguistics, and psychology. Working in active areas of the field usually involves learning the relevant background in these areas. Good introductory textbooks convey some of this background, but work on particular problems normally calls for reading more in related areas. Again, in general, the more you know, the better. For graduate work on many topics of artificial intelligence, expect to have to learn new areas of mathematics or other fields through independent study of texts and research papers. The following lists summarizes some of the typical level of knowledge useful in research.

• Artificial intelligence: The typical undergraduate course surveying search, logical representation, planning, and machine learning, or a solid class in database organization and semantics should be enough to get started, but expect to learn a lot more about logics, representations, and control theory.
• Mathematics: The typical undergraduate background in calculus, differential equations, linear algebra, combinatorics and probability really should be augmented with substantially more depth in logic, semantics, or statistics.
• Computation: The theory of computation, including basic complexity theory and programming language semantics, plays an important role in many investigations.
• Economics: Exposure to the basic elements of mathematical decision theory, social choice theory, and microeconomics (market theory) will be very helpful; they form central concepts in most of the topics of interest.
• Philosophy: Classes or reading in philosophy of mind, action, or philosophical logic will put you a step ahead in understanding the strengths and limitations of your work and the proper perspective on the mathematical and economic concepts.
• Physics: Basic knowledge of the character and content of modern classical and relativistic mechanics and quantum mechanics plays a role in understanding scientific applications, computability, and theories of limited rationality.

In particular, the areas of rationality and qualitative decision theory both involve substantial knowledge of mathematical decision theory and economics. The knowledge representation area involves substantial knowledge of mathematical logic, object-oriented programming, and database theory. The knowledge discovery area involves substantial knowledge of statistics, abstract algebra, and functional geometry in addition to knowledge of rationality and knowledge representation.

Writing a thesis and publishing articles on topics related to my areas of interest almost certainly requires knowing how to use LaTeX; see QC22 below.

A: There are three answers, each concerning a different sort of activity.

• Communication: Virtually all papers written to communicate or publish results in the relevant areas of research require use of LaTeX. Indeed, this is true for many parts of computer science and artificial intelligence. If all you know is Word or a similar WYSIWYG formatting system, you will need learn LaTeX as soon as possible, as I expect students to write summaries of problems and results throughout the process of investigation. The TeX Users Group has useful information about LaTeX and the TeX system on which it is based. The NCSU Statistics department teaches a course that provides useful references.
• Data analysis: The SAS data analysis software freely available to all NCSU students provides a good base for problems for which existing techniques suffice. However, if the analysis task at issue requires development of new ideas, new algorithms, or involves very large data sets, it probably calls for programming.
• Programming: I find Common Lisp the most convenient language for many purposes, and Java convenient for smaller scale efforts and for interface programming. My research systems have used both these languages for different components, HTML/XML for others, and C-family languages for some low-level tasks. I'm open to projects using any of these, as long as any necessary interface issues are handled responsibly.

QP1: What are my chances of admission, can the application fee be waived, do I really need GRE scores, ...?

A: These are all questions for the graduate administrator, and most of them are answered on the department web pages for graduate students. I will ignore email asking questions like this.

QP2: Will you recommend or secure my admission?

A: No. Admission decisions are made by the admissions committee, as a division of labor among faculty members. If we have never met, they have the same basis for decision as I would, and will undoubtedly ask me if they have any uncertainty about the matter. If you presented a paper at a conference I attended and we talked about your presentation afterward, let the committee know that you spoke with me. I will ignore emailed resumes, etc., sent to me by applicants.

QP3: Will you support me?

A: I do not provide financial support (i.e., research assistantships) to first-year students.

QP4: How do I get support?

A: Funding decisions for first-year students are made by the admissions committee of the department; contact the department directly!

It's a pretty good bet that the top-quality (upper 25-35%) graduate students will get a research assistantship after their first year, from some faculty member in the department. My best advice to brand new students is therefore to a) pick courses in areas in which you plan to specialize, and b) excel in your studies in your first year.

QP5: What should I do to get off to a good start at NC State?

A: See above.

QP6: Will you be my research advisor (i.e., may I join your research group)?

A: First, get admitted to the department. Second, enroll. Third, contact potential advisors during or after your second semester of graduate study. I will not respond to this question by email from students not yet enrolled at NC State.

QA1: How should I contact you?

A: Please use email to contact me. Please use only plain text email. Do not use HTML formatting or MIME encoding, which render the message difficult if not impossible to read.

I currently have to look through hundreds of pieces of junk email a day. If your email bears any similarity to such trash, or is not plainly readable at first glance, it stands a strong chance of being discarded without response.

QA2: How should I send you papers, information, etc.?

A: Please send me the URL by email, or ask me if I want to receive the document by email. Please do not send attachments unless I request you to do. If you find it necessary to send an attachment without asking me, send only plain text, DVI, or PDF files.

To avoid security problems and to increase the convenience of mail composition, I use a plain-text mail reader. Attachments require a cumbersome, multi-step process to transfer, decode, virus-check, and read. I am especially dubious about document formats designed so as to enable viruses and other malicious programs. I have a virus-checker, but don't always trust it. If I have any hesitation about a message, I might delete the attachment instead.

Please don't use an attachment to send me a simple text message. Copy the text into the message body instead.