C S C I     8 3 1 4


Projects

General

• Projects are due on Monday May 6th at 6:00pm You can bring projects to the last class or drop them in my office before the deadline set above. [Penatlies for late submissions.]

• Information and suggested projects Proj.pdf (PDF)

• This will be revised/expanded later with additional suggestions and references

What to submit

• You will need to submit a paper copy of your report. The paper will include sections such as (1) motivation; (2) methods used, relation to specific linear algebra problems seen in class; (3) experiments; and (4) optionally short computer codes. Page limit: no more than 12 pages. [Note~: this is not a hard limit - but try your best not to exceed it. Longer papers are not necessarily better]
  
  
• It is not essential (and not recommended) to include comprehensive (long) computer codes. The page length is not a criterion for grading (level completeness is important but length by itself is not -- see grading criteria below). Short matlab/octave scripts to show illustrations are OK.
  
  
• There is no specific format required as the projects vary a lot in their type and objectives (applications-oriented, survey, new algorithms; ...)

Grading Criteria

• The first criterion used is originality and overall intrinsic quality of the technical contents of the report. This boils down to the questions I ask myself after reading the report ``Is this an interesting report which can make it into a viable publication after some work''? ``Did I learn something in reading it?" ``Did the student show originality in his writing or did s/he just rush to put something together just to get some work done?"
  
  
• Clarity and overall presentation is the second criterion. It is absolutely essential that your report be geared to non-specialists Any symbol used must be defined. Any equation must be explained. A big turn off is when someone just puts equations from a paper and states the equations are .... If I do not understand the paper it will get a low grade. I highly recommend that you find a partner with whom you can exchange proof-reading of the reports [The partner will proof-read your report and you will proof-read his/hers. I will discuss this after the Spring break].
  
  
• The 3rd criterion I use for grading is based on the level of effort that seems to have been put toward the report. If you write a one-page report and think it is brilliant you will have to convince me in that one page that the content is indeed brilliant. If you write a 30 page report but clearly the effort to produce the report is minimal, it can be guessed and this will lower your grade. Remember that I expect the effort of a two-student project to correspond roughly to double the effort of a single-student project.
  
  
• The 4th and final criterion is relevance to the material taught in the course. It is not sufficient to say ``In the application I am dealing with, we need to solve a Riccatti equation. It is an important problem in control..." You will need to tie this to, e.g., other matrix equations and the QR and other factorizations, etc...
  
  
• In the past many reports received a good grade and virtually all received a grade exceeding 70%. I assign a specific grade for each of the above criteria and then add them. All 4 measures list above are equally weighted. So you can have a perfectly written and presented (25 points for second criterion) paper but if it not too relevant and not too original you may get a low grade.

How to get a good grade

• They illustrate some interesting behavior of known methods -- e.g., NMF and SVD for real life examples
• They show how some original method learned from the class has been deployed and successfully tested on a real research problem
• They present something to the reader that will provide a nice view of a given topic.. Here the originality is putting together a viewpoint, or a pedagogically well-crafted document.
• They implement something nontrivial from scratch and show real interest in the method developed --

How to get a poor grade

• Present poorly. If it takes me a long time to grade a report because I do not understand it, it will get a low score in ``presentation'' and possibly the other areas.
• Just discuss your own specialized research problem without much effort in adapting it to the need of the class.. There is a tendency to do a minimum if you have a paper ready (example: on something like target tracking and estimation) to just make minimal changes and give this as a term project report. Often this will get a low grade: Effort and originilaty will get a low score.
• Present a poor survey of a topic.. If you write a survey, it has to be reasonably exhaustive. Do a literature survey, present methods, compare pros and cons (according to what you read), ...