CS Honors Undergraduate Research Program

1
CS Honors Undergraduate Research Program

• Faculty Coordinator Prof. Amit Sahai
• Fall 2007 Spring 2008
• U.C.L.A.

2
What is the Honors Research Program?

• First of its kind at any University of California
  campus CS department
• Your own creative research Independent Original
  Work
• Under the supervision of excellent faculty and
  graduate student researchers

3
Honors Program

• Chance for you to shine in a creative way
• Cannot stress enough the importance of doing
  advanced independent work.
• The best way to distinguish yourself.
• Very impressive to top employers Google,
  startups even more so for grad schools.
• An excellent way to challenge yourself.

4
Outline

• What is research?
• The program
• Schedule
• Summary

5
What is Research?

• Formally advance state of art
• Informally tell people something new

6
What is Research and What is Not?

• Non-research
• My advisor gave me this mpeg decoding algorithm
• I learned about mpeg decoding
• I implemented it
• And it worked
• A lot of 199s are qualitatively like this

7
What is Research and What is Not?

• Research
• I took two existing mpeg decoders
• I took some sample movies
• I studied the decoders qualitatively
• I measured them quantitatively
• I concluded why one is better
• Why research
• analysis comparison something new

8
What is Research and What is Not?

• Research
• My advisor gave me this mpeg decoding algorithm
• I implemented it
• I measured it
• I analyzed it and found a bottleneck
• I instrumented the code to prove the hypothesis
• I recommend and conclude

9
What is Research and What is Not?

• Research
• I was given an mpeg decoding implementation
• I identified its bottleneck as above
• I proposed an improvement
• I implemented the improvement
• I measured it again to prove/disprove Im right
• I generalize and conclude

10
What is Research and What is Not?

• My advisor asked me to work on a big project that
  hes been working on with lots of graduate
  students.
• WARNING This is OK, but be sure to
• Have your own creative part of the project, for
  which you are primarily responsible
• Reasonably plan to spend no more than ½ of a
  quarter on initial part, where you build
  tools/codebase. Try to use tools and code that
  the team has already built.
• The most common way that a project can get off
  track is that you never get to the point where
  you analyze, experiment, suggest alternatives,
  etc.

11
What is Research and what is not ?

• Lots of possibilities For example, Building a
  web site
• How do you distinguish yourself from a high
  school kid writing a bunch of code ?
• A must make it novel. Something new or better
  than previous such websites
• How? Make it
• general can be created and configured from
  parameters and scripts
• automatically testable and demo-able
• a comparison between competing implementation
  technologies (different languages, databases, OS
  environments)
• a software engineering exercise in portability,
  robustness, performance, interface design,
• Use the stuff you learn in your CS classes !

12
What is Research and What is Not?

• Research
• Many possibilities
• So, what is research ?
• Formally advance state of art
• Informally tell people something new
• Not necessarily that much more work
• Just need to go the extra mile
• explore, analyze, generalize
• OK to get a negative result My idea
  didntwork, and heres why

13
What is research ?

• Research
• Analysis Synthesis Hypothesis
• Combination of work done before and new insights
• Literature survey understanding innovation

14
Other traits of a Good Project

• Interesting/important problem
• Non-trivial challenge(s)
• Exploration of new technology
• Can be finished in allotted time (3 quarters)
• Effective communication (talks, reports)

15
The Program

• Enroll in a special class CS 194
• Basic Elements of the Program
• Find an Advisor and a project
• Do the research!
• Checkpoints to keep you on track

16
Details
17
Checkpoints and Talks

• You will give 4 talks (3 for 2qtr) about your
  work, at different stages of the work.
• Purpose is not to give you busy work
• Main purpose To provide opportunities to
  re-evaluate and re-formulate your project and
  plan(Trust me, you will need to.)
• Secondary purpose To get practice presenting
  your work.

18
Schedule

• Sept 28 First meeting
• Oct 12 E-mail (to TAme) commitment from
  advisor, short description
• 1-2 paragraphs about project, 2qtr or 3qtr?
• Oct 15-19 - Project proposal presentations (to be
  scheduled through Google Calendar, TA Vipul Goyal
  will send out details)
• 5 min presentation of project idea (3-4 slides)
• 2 min presentation about project checkpoints

19
Schedule

• Guidelines about project plan
• For 2 qtr projects
• By end of first quarter, MUST have something to
  measure for projects involving software/hardware
  (most projects)
• For pure theory projects (very few), MUST have
  proved/disproved at least one conjecture you made
• For 3 qtr projects
• same as above, but by begin of 2nd quarter (i.e.
  Jan)

20
Schedule

• Nov 13-16 Arrange special meeting with advisor
  to discuss progress
• Ask the question -- am I on track? Do I need to
  scale down the project goals?
• Nov 16 -- Send an email with progress summary
  revisions to goals if any
• Dec 7 -- Checkpoint slides due (3-4 slides),
  should send to advisor, and upload (see TA for
  details)
• Jan 14-18 - Checkpoint talks - 5 min presentation
• SERIOUS review of goals, revision of goals

21
Schedule

• Feb 19-22 - 2qtr students Meet w/advisor to
  seriously review progress, last-minute changes to
  plan (including extend to 3qtr). Send email with
  progress summary revisions to goals if any.
• Mar 10-14 - Final talks (10 mins) for 2qtr
  students thesis due
• Mar 14 - Checkpoint2 slides (3-4) due for 3qtr
  students
• Apr 7-11 - Checkpont2 talks (5 mins)
• May 5-9 - (Same as Feb19-22)
• June 2-6 - Final talks (10 mins) thesis due

22
Find an Advisor and a Project

• In the first 2 weeks of the Fall quarter (start
  early)
• Get info about profs research
• Honors program page, home pages, research papers,
  word of mouth,
• Schedule meetings with several professors
• email, office hours, appointments
• If having trouble, try to catch prof just after
  their class
• Warning some profs are not around. Have backups!
• Choose a professor
• Should be from CS dept
• Can work with someone outside CS, If youd like
  to do this, consult with me
• Can be jointly advised by multiple profs

23
Find an Advisor and a Project

• Decide on a project
• Profs suggest choices
• Students come up with their own
• A combination
• Mutual agreement, interest, enthusiasm
• Write brief description of project and get
  Advisors email commitment to advise you

24
Find an Advisor and a Project

• Topics/areas that may not be obvious research
  areas of profs
• Games and game playing
• Education aids
• Language recognition/translation
• Wireless
• Cross-discipline (econ, history, math, psych,
  politics, sociology, etc.)
• See http//honors-program.wikispaces.com/

25
Project Proposal Talk

• Problem description
• What am I going to do?
• Why is it important?
• Why is it hard?
• Approach
• Previous approaches
• My approach
• Why is mine better?

26
Project Proposal Talk

• Methodology, milestones, deliverables
• Plan of attack
• Specific steps
• What steps/deliverables will be done by
  checkpoint (end of Winter quarter)
• What other steps/deliverables will be done by end
  of projecct (end of Spring quarter)
• What might be hard and whats the fall-back plan
• Summary

27
Project Proposal Talk

• Dont have to talk about everything
• But include everything (in notes section or
  other places) in the slides
• Be specific, give details of plan
• Tell me whats the new/clever/cool nugget
• Proposal talk is not your starting point some
  preliminary work should have gone into the
  project by then (i.e. in the next 2 weeks!)

28
Project Proposal Talk and Beyond

• Scope of Project
• Not too little
• Not too much (carve out a piece, limit
  functionality, reduce measurements)
• If youre ambitious, have a longer term plan but
  the short term plan should still be doable
• Dont be afraid of getting negative results
• Have intermediate results

29
Project Proposal Talk and Beyond

• Be conscientious
• Start early
• Define small milestones for yourself
• Work continuously to meet milestones
• Meet with your advisor regularly
• Dont hesitate to get help

30
Project Checkpoints and Talks

• 3-4 slides
• What you proposed to have done by checkpoint
• What you have actually accomplished by checkpoint
• Steps
• Deliverables
• Difficulties/surprises/deviations ?
• What more do you expect to do
• Steps
• Deliverables

31
Final Project Results Talk

• Review the problem description and proposed
  approach give the theme
• Give details (e.g., of implementation) to support
  the theme
• Give key results to support the theme
• Summarize the theme

32
Final Project Thesis

• No minimum length.
• Introduction
• Background
• Problem description include goal
• Approach
• Previous approach(es)
• My approach
• Why is mine better
• Detailed description of methodology or
  implementation

33
Thesis (cont.)

• Experimental results
• Analyze/interpret data, dont just give numbers
• What does this have to do with your theme?
• Conclusion
• Acknowledgements and bibliography

34
Credits

• Thanks to the folks in charge of the Princeton CS
  Independent Research Program (especially Moses
  Charikar and Randy Wang) for most of the
  (interesting) material in this talk.