CS786 - Assignments

There will be five assignments, each worth 8% of the final mark.  Each assignment must be done individually (i.e., no team) and will consist entirely of programming questions.  More precisely, you will be asked to program some algorithms for probabilistic inference and machine learning and to test them on some datasets. 

Assignment 3: due July 30

• Handout
• Skeleton code:
  
• skeletonJuly19.jar: skeleton without any public tests
• skeletonJuly23.jar: skeleton with 21 public tests and all necessary datasets (*.csv) for the ublic and release tests.  This is the same skeleton code as released on July 19, except for thePublicTests.java file and several *.csv datasets. If you already started writing your code in the skeleton released on July 19, you can continue with the same files, but make sure to copy the new/modified files: PublicTests.java, *.csv.
• Marmoset:  5 projects labeled 3a, 3b, 3c, 3d and 3e were created in Marmoset to allow you to test each function separately.  Hence, once you pass the public tests for a function, you can immediately see whether you pass the release tests for that function instead of having to wait till you pass the public tests of all functions.
  
• Grade: # of release tests passed There are 10 release tests in total.
  

Assignment 2: due July 9 July 13

• Handout
• Skeleton code:
• skeletonJune26.jar: skeleton with only 4 public tests
• skeletonJuly2.jar: skeleton code with 20 public tests and several new networks and datasets.  This is the same skeleton code as released on June 26, except for the PublicTests.java file which includes many more tests and several new .bn, .mn and .csv files that are necessary for the new tests.  If you already started writing your code in the skeleton released on June 26, you can continue with the same files, but make sure to copy the new/modified files: PublicTests.java, *.bn, *.mn and *.csv.
• skeletonJuly9.jar: skeleton code with the necessary datasets for the release tests.  This is the same skeleton code as before, with new .bn, .mn and .csv files that are necessary for the release tests.  If you already started writing your code in an earlier skeleton, you can continue with the same files, but make sure to copy all .bn, .mn and .csv files. 
• Grade: # of public and release tests passed.  There are 20 public tests and 10 release tests, hence your grade will be out of 30 for this assignment.

Assignment 1: due June 6 June 17

• Handout
• Skeleton code:
  
• skeletonMay30.jar: skeleton code with only 2 public tests, 1 Bayes net and 1 Markov net
  
• skeletonJune4.jar: skeleton code with 48 public tests, 3 Bayes nets and 2 Markov nets.  This is the same skeleton code as released on May 30, except for the PublicTests.java file which includes many more tests and 3 new networks (loopy.bn, polytree.bn and grid.bn) that are necessary for the new tests.  If you already started writing your code in the skeleton released on May 30, you can continue with the same files, but make sure to copy the new/modified files: PublicTests.java, loopy.bn, polytree.bn and grid.mn.
• skeletonJune8.jar: the file polytree.bn had a typo in the sense that the conditional probabilities for variable I (i.e., Pr(I|E)) did not sum to 1.  This new skeleton jar includes a corrected version of polytree and a corrected version of the associated PublictTests.  If you already started working with an earlier version of the skeleton, simply copy the files polytree.bn and PublicTests.java.  Make sure to include the corrected version of polytree.bn in the jar file that you upload to Marmoset.
  
• Marmoset submit server: https://marmoset.student.cs.uwaterloo.ca
• Make sure to login Marmoset with the first 8 characters of your UW userid (userids longer than 8 characters are not supported).
• Submit your code by uploading a jar/zip file that contains the same files as in skeletonDate.jar.  To practice, you can upload the skeletonDate.jar file provided on this webpage.  Marmoset will simply indicate that it compiles, but none of the public tests are pasts.  You can upload submissions as often as you wish until the deadline.  Your best submission based on the release tests will be used for grading.
  
• Once a submission is uploaded, wait a minute or two and refresh your browser.  The number of public tests passed will be displayed.  You can click on "view" to see which public tests failed and why.
• It is possible that the public tests reject correct solutions simply because I didn't think about all the possibilities when coding the verification routines.  If your code fails a test, but you think that your solution is correct, send an email to Pascal to double check.
  
• Release tests:
• 18 release tests are now uploaded on Marmoset.  You don't have to do anything to run previous submissions against the release tests since Marmoset automatically re-evaluated all submissions as soon as I uploaded the release tests.
• You can request to see your results for the release tests only after all public tests are passed.  The rational is that you should be debugging your code with the public tests before checking your results for the release tests.  Once you have a submission that passes all public tests, click on "view" for that submission and scroll to the bottom of the screen.  Click on "release test this submission" to see the results of this submission for the release tests.  
• Grade: # of public and release tests passed.  There are 48 public tests and 18 release tests, hence your grade will be out of 66 for this assignment.