Final presentations will consist of a paper (If you are doing the review paper option OR a talk if you are doing the project option). If you wish to do both, or switch (talk for review paper, paper for project) please talk to us about it. Talk presentations can be live or video (or both). ----------- Guidelines for time frames: 10 minutes for 1 person groups 12 for 2 people, 15 for 3-4 and 17 for 5-6, 18 for more than 6. ----------- Components of a project talk: Introduction and Motivation -- What were you trying to do and WHY? - For full points clearly state the problem and why it is important. Related work -- Mention any related work from our class readings. For full credit also mention a recent paper in the literature or commercial product that addresses this problem (or if there is nothing like it in the literature or market make the case for why your problem is unique). Methods - What did you do? Be as precise as you can. For full credit there should be enough detail for someone relatively skilled to replicate your work. Results - What did you discover? How well did it work? As this is a class project, it is likely that many things did not work as well as planned. For this project, detailing what went wrong is as important as describing what went well. Discussion - What did you learn? What could you do better? (What would you have done next if you had more time)?..... Why do you think it didn't work if it didn't? If everything worked perfectly, what next steps would you suggest for followup development. For full credit discuss two extensions or improvements to your project with short justifications for why you think that would work better (improvements) or why they are promising extensions. ------------------------------------------------------------------------- Components of a review paper (5-8 pages): Introduction and Motivation -- What were you interested in researching and why? Related Work - Mention any related work from the class readings/slides. Review of your Main Literature -- What were the studies you read about? What were the findings? Why are these papers interesting? (Why did you select these papers?) How are the different papers/algorithms related to each other. (If you just read one paper/algorithm - then you would describe in your own words how the algorithm works and add your own insight about how it is related to other algorithms you know or could have chosen (from class material or other material you have encountered or looked at before choosing your topic) Discussion - Any overall conclusions/summary -- How are the findings related? What studies would you suggest to do next? What is your opinion of the strengths and weaknesses of the algorithm/idea(s). Here is a link to a survey/review paper that is much much longer more thorough than yours should be, but at least is an example of a paper in this style. (Your paper should be about 5-8 pages) http://scholar.google.com/scholar?q=Data%20Clustering%3A%2050%20Years%20Beyond%20K-Means%20author%3AJain (top link) There are a lot here at http://mlsurveys.com --------------------------------- Grading Considerations First -- all projects with good effort will receive good grades, it is not necessary to have a working successful algorithm or demo. The most important thing is to demonstrate what you learned and that you thought about the problems and issues even if you couldn't solve them all. (Note you can ask us questions to help with this deep thinking also -- through piazza or by emailing for a meeting) We are looking for (so try to show in your presentation (written or oral): 1) Do you know why you are doing the project? 2) Can you relate it to work from the class? 3) How hard was what you did/tried to do? (here we recognize that some projects were hard just to get the data uploaded/transferred in useable form, whereas for others that is not an issue) 4) What did you learn? 5) How much work did you do? (also here we recognize that some projects were hard just to get the data uploaded/transferred in useable form, whereas for others that is not an issue. Also here there might be consideration for how many people were in the group though we recognize that 3 person groups can't do 3 times the work, they should do more work than one person) Helping other groups (directly or through piazza posts) would receive extra credit here also. 6) Can you clearly present the motivation and what you did/tried to do to the class? 7) Did you get something to work (this is a bonus -- not required)? 8) Did you think deeply about the results or about problems with the algorithm or approach? 9) Are you aware of the strengths and limitations of your approach/algorithm? 10) How much do you understand about the problem, algorithm, and approach?