Sample PosterPosters are due to be submitted by Thursday (05/03) midnight in a pdf format on canvas. You can find some of the sample posters here.
- Projects proposals are due on March 3 2018.
- Within a week we will give you comments on your proposal or a request to modify/augment/do a different project.
- Any project that has a significant Machine Learning component is good.
- You may do experimental work, theoretical work, a combination of both or a critical survey of results in some specialized topic.
- If you do experimental work, we expect more than: we took this data; we compared multiple learning algorithms on it. As a minimum, you should focus on a specific problem/issue and analyze it – this could be some feature analysis, adaptation to new domains, or any other hypothesis that might be appropriate.
- The work has to include some reading. Even if you do not do a survey, you must read (at least) two related papers or book chapters and relate your work to it. Include citations already in your proposal (do it in a way it is done in scientific articles.)
- The project proposal should clearly state, in 1-2 pages that are formatted like a paper, the objective of the project, how you intend to go about it, reading list, and what you have already done. Naturally, it should have a title and include the names of the participants/authors.
- Originality is not mandatory but is encouraged.
- Try to make it interesting!
- Notice that due to the size of the class, we will not accept individual projects. Projects should be done in teams of size 2-3.
Eventually, you will write a short paper (4-6 pages; 11 font; please use the style suggested here). We are planning a poster session for all the projects (that will likely happen during the time scheduled for the final exam).
Below are guidelines on how to write up your report for the final project. These are only guidelines; you will need to adjust it to the problem you are investigating, but try to structure your report along these suggestions and make it look like an article. Please don't use the guidelines below as subtitles of your paper; these are just guidelines. Try to make it look like a published article.
- Introduction: Motivate and abstractly describe the problem you are addressing and how you are addressing it. What is the problem? Why is it important? What is your approach? What is the goal of your paper? Provide a short discussion of how it fits into related work in the area. Summarize the basic results, conclusions and contributions that you will present. All these apply equally to experimental papers, survey papers or theoretical papers.
- Problem Definition and Algorithms
- Task Definition: Introduce the model and/or problem you are studying and define the notation you are going to use. Precisely define the problem you are addressing (e.g., formally specify the inputs and outputs). Elaborate on why this is an interesting and important problem.
- Algorithm(s) Definition: If you study learning algorithm(s) experimentally this is the place to present it. Describe in reasonable details the algorithm(s) you are using. A pseudo-code description of the algorithm you are using is often useful. Depending on the context, it may be useful to trace through a concrete example, showing how your algorithm processes this example.
- Expectations: In case of an experimental study, discuss what you hope to achieve. How do you expect each algorithm to behave and why. Try to justify your hypothesis as rigorously as possible. Discuss how your expectations drive your experimental design.
- Experimental Evaluation
- Methodology: What are the criteria you are using to evaluate your method? Describe the experimental methodology that you used. What is the training/test data that was used, and why is it realistic or interesting? What performance data did you collect and how are you presenting and analyzing it?
- Results: Present the quantitative results of your experiments. Compare to the literature (state-of-the-art, ideally) and to an sensible baseline method. Graphical data presentation such as graphs and histograms are often better than tables. What are the basic differences revealed in the data. Are they statistically significant?
- Discussion: Is your hypothesis supported? What conclusions do the results support about the strengths and weaknesses of your method compared to other methods? How can the results be explained in terms of the underlying properties of the algorithm and/or the data.
- Data Analysis: it is often useful to discuss error and/or analyze some test cases as a way to illustrate the problem, your method, the limitations, etc.