You cannot sign up to Kaggle from multiple accounts and therefore you cannot submit from multiple accounts.
\r\n\r\n Privately sharing code or data outside of teams is not permitted.\r\n It's okay to share code if made available to all participants on the forums.\r\n
\r\nTeam mergers are not allowed in this competition.
\r\n\r\nThere is no maximum team size.
\r\nYou may submit a maximum of 2 entries per day.
\r\nYou may select up to 2 final submissions for judging.
\r\nStart Date: TBA
\r\nMerger Deadline: None
\r\nEntry Deadline: None
\r\nEnd Date: 2/1/2014 12:00 AM UTC
\r\n\t\r\nThe rules page is a bit special: it will auto-update the text based on the parameters of the competition. The rules you need to add here are any additional rules relating to dataset usage. You may also put grading information here.
", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "data-description", "content": "This is where you find the dataset as described in the overview
", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Description", "content": "\n\nThe year is 1050 BC near the warm beaches of the Mediterranean Sea. The Martians have invaded. It is a time when earthlings have to unite against the colonial Martians who are conspiring against us to take over the planet. The Martians can time travel too! They travel to the year 1997 to a small town in the United Kingdom and pick up three things, an English Grammar book, Good-Turing’s paper, and newspapers from a local market. Their aim is to destroy Earth retroactively. They come back to the year 1050 BC and alter everything in Earthling’s civilizations history to make them never land on Mars, their beloved land. \nModern day language and the alphabet was in the formative stages in 1050 BC. The Phoenician alphabet is the oldest verified consonantal alphabet. The Martians now have a diabolical plan. They want to spread nonsensical news amongst the Phoenicians for them to believe this nonsensical news is valid language. If language had never developed, do you think humans could have ever sent out missions to Mars? \nThe one thing Martians are good at is statistics (They do have Good-Turing’s paper after all). They develop a trigram language model using the GT paper, which they will use to generate nonsensical news in the Phoenician alphabet using the newspapers they picked up from the year 1997. You are a local in Greece, and an independent researcher, who wants to stop the martians from succeeding by removing all the fake articles from circulation. Your job is to distinguish between the fake and real articles that are being circulated in your town by using your statistics knowledge which Prof. Bhiksha has taught you in your dreams. (Well, CMU 11-761 Fall 2018 course has a dream edition too).\n
\n\nHere are the files at your disposal:
\nThe following files can be found in the data section of kaggle:\nThe collection of newpapers the Martians took from 1997 in the Phoenician alphabet. You may use this to build your own language models.
| Number of sentences | \nNumber of articles with this number of sentences | \n
| 1 | \n20 | \n
| 2 | \n10 | \n
| 3 | \n10 | \n
| 4 | \n10 | \n
| 5 | \n10 | \n
| 7 | \n10 | \n
| 10 | \n10 | \n
| 15 | \n10 | \n
| 20 | \n10 | \n
[Note: the set on which you will be tested (\"test\n set\") will have the same format, approximately the same number of real and fake articles, and the same length distributions as this development set. \n Of course, the shorter the truncated article, the harder the classification\n decision.]
The format is the same as in trainingSet and developmentSet. You are required to submit your final predictions on this testSet. Your output file must contain 200 lines with predictions (0/1) \n", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Evaluation", "content": "
There are 2 evaluation metrics Hard and soft metric. We will be evaluating the hard metric on kaggle. For this, for each article in the test set, you will output the first line as id,category (no whitespace after comma) and the subsequent lines in that format, where id is the document number indexed from 1 and category is 0/1 (You will submit only this file to Kaggle). For the soft metric: For each article in the test set, your program should output the two\nposteriors: P(fake|article) and P(true|article) (which of course should sum to\none), in that order. It should also output a classification decision: 0\nfor \"fake\", 1 for \"true\". Typically (but not\nnecessarily), the latter would be 1 if P(true|article)>0.5, 0 otherwise, with\nsome tie-breaking for the 0.5 case. These three numbers should be output\non the same line, separated by a single space, one line per article, e.g.:
\n0.791243 0.208757 0\n0.300781 0.699219 1
The performance of your program on the test set will be judged in the\nfollowing two ways:
\n 1. \"Hard\" metric: % correct classification
\n 2. \"Soft\" metric: average log posterior of the\ncorrect label [ P(correct_label|article) ]. (You won't know the correct\nlabel of the test set, but the grading program will). This is akin to the\nlog perplexity of predicting the \"language\" consisting of the two\n\"words\", TRUE and FAKE. It will be wise to avoid giving your prediction a probability of 1, unless you are absolutely sure (think about what will happen it you are actually wrong.)
You are free to use the given files in any way you wish. One reasonable\napproach would be to study the differences between the true and fake articles in\nthe 1000 article trainingSet, using both manual \"detective\nwork\" and automatic statistical analysis, and use this to suggest various\naspects along which the two differ, optimizing parameters on this training\nset. The 200 article developmentSet can be used to (1) test and\nvalidate suggested improvements, and perhaps to (2) optimize a few additional\nparameters. You would also use the development set to (3) estimate the\nexpected performance of your program on new data. Note though that the\nmore you use the development set for (1) and (2), the more your solution will\nadjust itself to that set, and the less accurate would be the estimate derived\nin (3).
\nYou are encouraged to use any and every statistical technique or concept that\nwe learned in class (and others too). And, perhaps more importantly, you\nare encouraged to use anything you learned about the properties of language (and\nof trigram models...). You are free to use external resources.
\nFor this project you will form groups of exactly three people.\n You are required to form your team no later than Nov 10th, and you must email Pramati (pkalwad@andrew.cmu.edu) your group members by then . One email per team, with other team members cced. In your report, please clearly state who did what in this project.
\n\n 1. Performance of your program, on both hard and soft\nmetrics. Hard metric will be calculated on Kaggle, soft metric will be done after the competition is over by the TAs.
\n 2. The extent to which you exhibited understanding of\nstatistical techniques and sound statistical methodology.
\n 3. The extent to which you exhibited understanding of the\nproperties of language.
\n 4. Creativity and elegance of your solution.
\n 1. A tar.gz or zip archive containing your code, ready to run.\n
\n You can develop wherever you wish, but it must compile and run on Andrew's UNIX login servers so the TAs can test your code.
\n The code should be invocable by\nissuing the command ./RunMe in that directory (RunMe could be an\nexecutable or a shell script).
\n If your program makes use of any\nfiles, they must reside in that same directory, and be accessed via hard-coded\nfilenames.
\n Your program should take its input\n(in the format of developmentSet) from stdin, and write its output to\nstdout. We will run your program with the command:
\n ./RunMe < testSet.dat > testOutput.dat
\n 2. The performance measures of your program on the\nDevelopment set (both metrics).
\n 3. A clear but concise write-up detailing your work,\nincluding all the ideas you tried (even those that didn't work!) and their\nimpact.
\n 4. A clear description of what each group member\ncontributed.
\n 5. Comments on this assignment and on your experience with\nit, and suggestions for improvements.
\nPlease name your report pdf file in the form of FirstName_LastName_project_report.pdf and submit it via Canvas. Remember to also submit your code as FirstName_LastName_project_code.zip\n
\nProject presentation will be on Dec 3rd and Dec 5th. Half of the teams will present on Dec 3rd, and the other half on Dec, the 5th. The date assignment and presentation order will be decided by lottery, and in order not to discriminate, will only be announced on Dec 3rd 11:30 am -- 30 minutes before the first presentation. Each group has exactly 15 minutes for a presentation. There will be Q&A (not included in the 15 minutes). Each team member should have approximately same amount of time to present his/her part of work. You will not be graded individually on your presentation.
\n\nPlease prepare the slides for your in-class presentation. You must submit your code, report and your slides for your in-class presentation by ***Sunday Dec 2nd, 11:59 pm***.
\n\n