General Advice on CSE Research* Bharat Jayaraman CSE Department University at Buffalo * compiled from many sources

Outline of Talk CSE Department Research Faculty Research Activities Doctoral Research Research Strategies Qualities of a Researcher Online Resources

What is Research Systematic Study or Inquiry Leading to new knowledge or theory, or creation of a new artifact, process, etc. Goal of research is to advance a field, be it science, engineering, or the arts.

1. CSE Department Research Core Areas: Algorithms & Theory Artificial Intelligence Systems (SW/HW) Lots of multidisciplinary interests

Multidisciplinary Research Other Engineering Departments Bioinformatics Pharmacy Public Health Math, Chemistry, Biology, Geography CSE Media Studies, Communication Management, Law Linguistics, Philosophy Psychology, Anthropology

2. Faculty Activities Teaching Research Service

Faculty Research Activities Perform research (more on this later) Publish journal and conference papers Speak at conferences and other institutions Secure external research funding Manage research groups

Faculty Service Activities Serve as reviewer of papers, proposals Serve on conference committees Serve on journal editorial boards Serve on national boards and panels

3. Doctoral Research Develop an interest in an area, by finding out what are the major open problems; what are the major research groups; what are the major conferences and journals Identify and meet regularly with research advisor Take advanced courses, build technical strength Attend and present at research group meetings Write and submit papers to conferences/journals Attend conferences and workshops.

Major Steps for PhD in CSE Qualifying Process (coursework) Major Professor Selection Dissertation Committee Dissertation Research Proposal Dissertation Completion Dissertation Defense Ph.D. degree

Outline of Research Proposal Problem and its Significance Related Research Proposed Approach to Problem Results Obtained to Date Remaining Research Expected Contributions

Good Research Problem Important (and interesting) Related to other problems Not easily solved Can be solved Can be solved by you Leads to new research

Some Grand Challenges P = NP? Machines that learn Secure systems Ultra-fast computers … Grand challenges provide high-level motivation for one’s research. They often require multidisciplinary approach.

4. Research Strategies Motivation is “top-down” Execution is “bottom-up” Identify specific problem Study related work thoroughly Must be able to present your work in the light of what is known, highlighting your contributions

Literature Search Read papers from the top conferences, journals, and researchers Read “actively” and identify possible areas of further work Write a para on each paper giving the main points – this may go into thesis Consider writing a survey paper from the literature search

The Hourglass Model* * http://www.socialresearchmethods.net/kb/strucres.htm

When a problem is very hard Formulate a simpler problem, solve it, and write it up – this may be substantial enough for a first publication! Then try to extend solution. E.g. The difficulty of general program verification in ‘70’s led to research on model checking of finite-state systems in ‘80’s, which became practical for hardware verification.

Cross Fertilization Techniques in one area often have applicability in other areas. E.g. Hardware Verification via Software Testing; Email Spam Detection via Machine Learning; Program Debugging using Temporal Databases. National Science Foundation has funding for “Cross-Cutting Programs”

Details are important Getting the fine details correct is extremely important in research. Sometimes, this can make or break a major finding, especially a theorem. In 2010, there was a purported proof that P != NP. But upon inspection of the details, it was found to be wrong.

“Negative results” are useful It takes time for ideas to crystallize. Sometimes, the search for the right solution involves going down unproductive paths. Be on the look-out for “negative” results, i.e., proving that a problem cannot be solved is also useful.

5. Qualities of a Good Researcher Logical Thinking and Ability to Abstract Creative, Innovative Attentive to Details Well-informed, has broad background Communication Skills Passionate Perseverance Take success/failure in stride

Qualities of a Good Advisor Has research interests in common with you Has a national or international reputation Has directed students in the past Has a reputation as fair and reasonable Has an active research group Has grant support for research Is someone you like and admire

6. Research Resources Research Advisor Texts, research monographs Journals, conference proceedings How to improve one’s writing Many online resources available today.

Parable of the Black Belt After years of hard work, the student appears before his master who asks one final question: "What is the true meaning of the black belt?“ "The end of my journey," says the student. "A well-deserved reward for all my hard work." “You are not ready for the black belt,” says the master, “return after one year!”

Parable of the Black Belt After another year of hard work, the student appears before his master who again asks: "What is the true meaning of the black belt?“ "A symbol of distinction,” says the student, “the highest achievement in our art." “You are not ready for the black belt,” says the master, “return after one year!”

Parable of the Black Belt After another year of hard work, the student appears before his master who again asks the same question. Now the student replies: "The black belt is just the start of a never-ending journey of discipline, hard work, and the pursuit of an ever-higher standard." “You are now ready for the black belt,” says the master, “go and get started on your work!”

Notes for Future Use Knowledge Tree – where are you extending the tree 2. Math/Science/Engg – what type of research and approaches within each 3. When you have difficulty reading a paper – what to do.