Your paper should be 4-6 pages, and should have the following main sections:

  1. Abstract
    The abstract is a brief summary of your work. It should be written to make the reader want to read the rest of your paper; think of it as your "elevator speech". Briefly state the basic contents and conclusions of your paper: the problem you are addressing, why the reader should care about this problem, the algorithms/methods for approaching this problem, and the main results and/or contributions of your work. Limit your abstract to 200 words at most.

  2. Introduction
    The introduction is the big picture of your work: what, why, and how. It includes a definition of the problem you are solving, a description of why that problem is worth solving, and a high-level description of the solution(s) you apply to the problem. In addition, it motivates the work you are doing (i.e. why should a reader find your work important), and describes your contribution to the area (this may not be applicable to your project).

    Remember that the target audience is a classmate who knows about the topics covered in the course but not your particular research question. Your introduction should successfully convey the culmination of your 4 weeks of work. It should place your work in the context of both the course and related work in the field.

  3. One or more sections describing your approach(es) This is some times referred to as Methods, Algorithm, Methodology, or Approach. Your section (or sub-section) headings could even be specific to the algorithm, such as k-Nearest Neighbors. As a whole, these sections should tell the reader what you did, and how they could replicate it (i.e. someone with your level of background before starting the project should be able to reproduce the results of your project based on this description). Note that this section indicates how deep you investigated a problem, and mimics the presentation of methods as done in class. Be sure to address these issues:

    • Details of each algorithm and method you applied to the problem, at a algorithmic level. That is, tell us what a support vector machine is before telling us how you applied it to your data.

    • Your methods should include figures and/or pseudocode to help illustrate the approach. This could be a pipeline if you synthesize many approaches, toy example to explain a complex model, pseudocode (search for latex packages that fit your need), graphical models, and more.

    • Analysis of why these methods may be suitable.

      Be sure to discuss with us any questions you have about how this applies to your particular project. It is very important that this section is done in your own words - do not lift descriptions and images from other papers or online resources (except when citing very specific facts).

  4. Experimental Results

    • Experimental Methodology: Explain how you gathered the data and details of how your experiments were run. Did you use any external software? Did you pre-process the data in any way? How did you generate features from the data? Did you use cross-validation? Do you need to define any questions you use for evaluation (e.g., Silhouette Index).

    • Explain the tests you performed (and why)

    • Present your Results
      Choose quality over quantity; the reader will not be impressed with pages and pages of graphs and tables, instead s/he wants to be convinced that your results show something interesting and that your experiments support your conclusions. Your results should be rich in meaning and concise.

    • Discussion of your results.
      Explain/interpret your results (possibly compare your results to related work). Do not just present data and leave it up to the reader to infer what the data show and why they are interesting. What conclusions can you draw?

  5. Social Implications In a couple of paragraphs, identify the key stakeholders and the impact your project could have. This section will vary quite a bit between each project, so please talk to us if you need some guidance. You should be thinking about questions such as:

    • Who are the key stakeholders?

    • Who stands to benefit from this work, and how? Are there groups that may suffer? There may not be a stark dichotomoy; that is, it may be clear that there will be implications for certain groups, but it isn’t obvious of those will be a net benfit or not.

    • How does this work interact with existing group dynamics? Does empower stakeholders, or disempower them? Does it shift power from one group to another?

    • Are there concerns on how this may be used in a real-world scenario?

    • You are encouraged to use this section to connect to your expertise in other fields e.g., legal, economic, ethical, etc. implications.

    • Avoid picking one extreme or the other, we are looking for you to identify tensions rather than gives a sales pitch for your project (or alternatively, condemn the work completely).

  6. Conclusions & Future Directions for your work Conclude with the main ideas and results of your work. Discuss ways in which your project could be extended — What’s next? What did you want to try but not have time for? What are the interesting problems and questions that resulted from your work?

  7. References
    At the end of your paper is a Reference section. You must cite each paper that you have referenced — your work is related to some prior work. You must cite all referenced papers within the text of your paper.

Writing Style Guidelines

  1. Write in a top-down style
    First present the high-level issues, then expand them. This applies to the overall organization of your paper as well as the organization of sub-sections and individual paragraphs.

  2. Tell a coherent story
    Your paper overall should hold together; each piece should be a necessary part of the whole. Content should not exist merely for the sake of taking up space, but rather to convey important information to the reader (e.g. only explain SVMs if later sections require the reader to understand SVMs). Paragraphs within each section should flow appropriately.

  3. Be concise and precise
    This is not an essay for an English class; use as many words as you need to convey a correct and truthful understanding to the reader, and no more. Avoid flowery language, and do not use hyperbole.

  4. Conclude each paragraph, section and entire paper
    Each chunk of your paper whether it be a paragraph, a sub-section, a section, or the entire paper should have a conclusion. For example, each paragraph should be written as follow:

    • 1st sentence(s): main idea of paragraph

    • middle sentences: expansion of the idea (further explanation or elaboration of the topic)

    • concluding sentence(s)

Each section of your paper should be organized as: high-level important points first, details second, summarize high-level points last.

  • Write clearly and understandably: tone, style, grammar and spelling are important in making your work readable; a paper that is difficult to read and interpret discourages the continued interest of the reader.

  • Use active 3rd person as a default: "We present", "we show", "we demonstrate"…​

  • Use passive 3rd person for things you did in the past: "we ran k-fold cross-validation"…​

  • Define terms and acronyms, and always define them before using them

  • Don’t shy away from math: use equations and variables instead of overly verbose descriptions.

  • Use figures: Use diagrams to help explain system design, and graphs or tables for presenting results. The best papers have figures to help introduce the paper and problem, methods, and results.