COSC 388: Machine Learning

Paper Guidelines

Due: Dec 5, 5 PM

The final paper should describe what you have done and how it worked. It should also speculate on why your approach did or did not work and on directions for future research. Here is a sample outline for a paper describing an experimental study:

  1. Abstract
  2. Introduction
  3. Statement of the Problem
  4. Literature Survey
  5. Approach
  6. Experimental Study
    1. Method
    2. Results
    3. Analysis
  7. Conclusion (of the paper)
    1. Summary of the paper
    2. Inferences drawn from your findings
    3. Future Work
  8. References
Some of these sections will need to be renamed or will be irrelevant, depending on the type of project you undertake. Furthermore, you might want to collapse the section on the problem statement into the introduction. In the conclusion, the summary, inferences, and future work do not necessarily need to be separate sections; they also can be paragraphs. Nevertheless, you generally need to address the following questions:
  1. What specific problem did you address?
  2. Why is this problem important for the field or for society?
  3. How have others attempted to address this problem?
  4. How have these attempts fallen short?
  5. How did approach the problem?
  6. What is your research hypothesis?
  7. What experiment(s) did you conduct to support your hypothesis?
    1. What algorithms/methods/systems did you use?
    2. What data set(s) did you use?
    3. How did evaluate your system?
    4. What metrics did you use to measure performance?

You should pick a convention for citing material. I honestly don't care which one you pick as long as it is not of your own making and you use it consistently and correctly. The Chicago Manual of Style and the Publication Manual of the American Psychological Association both contain information about how to properly cite and reference scholarly material.

Not all references are created equal. Peer-reviewed journal and conference papers are the gold standard, so most of your references should come from these sources. Books, edited volumes, and invited papers are also important and acceptable. Newspaper and magazine articles are usually not acceptable except to motivate the need for research in a certain area. For example, a Post article on the number of computer break-ins at the Pentagon could be used to motivate research on machine learning approaches for computer intrusion detection.

Furthermore, in the age of the Web, you must be very careful about using material from Web pages, simply because there are no checks and balances, no filters for credibility and veracity, for what someone can place there. The Web is an important research tool, but one that does not yet supplant traditional library research. Students in the past have used material they found on the Web as if it came from a peer-reviewed source. If you're unsure about a reference, bring it by, and we'll take a look at it.

The format and length---and quality---of the paper should be comparable to a workshop or conference paper: 6 pages, single-spaced, 10-12 point type, 1 inch margins all around. Here are samples in PDF and PostScript formats: [sample.pdf | sample.ps]


Go Back Send Mail