Assignments

The two assignments are carried out individually and are examined orally.

• Assignment 1 (Deadline November 19)
  • Code for index in Part 3
• Assignment 2 (Deadline December 16)   
  
  • Data for ICA in Task 2.4: DataICA.txt
  • Data for LDA in Task 2.5: R3_all_Dictionary.txt, R3_GT.txt, R3_Label.txt, R3-trn-all.txt, R3-trn-all_run.txt, R3-tst-all.txt, R3-tst-all_run.txt
  • Tips for Task 2.6:
    • Gibbs sampling is generally very slow. You can expect one training of the LDA model with all training data to take several hours. Therefore, plan ahead, and start working on this assignment as soon as possible (today!) if you want to do it. You can work on other tasks while you wait for it to run.
    • Set both Dirichlet hyperparameters to 0.3, this is a good value. You can consider it given, and do not have to motivate it in the report.
    • Convergence takes in the order of 100 iterations, but print the change in z between iterations to track the convergence.
    • First make sure that your code is correct by training the model with just a part of the data - 1/3rd or 1/4th. But in the report you should show a training result with all data.
    • Since the z:s are randomly initiated, you will get slightly different convergence times. Start off several training sessions at the same time, to increase experimental efficiency.

Grading

Each assignment is graded as A-F(fail). In the case of F, the assignment has to be submitted again at a later time. The grade depends on how large part of the assigment that was carried out correctly - this is specified on each assigment description. Assignments that are handed in after the respective deadline receive a maximum grade of E.

In other words: Be on time with the assignments! 

Late assingments are handed in before 12.00 (NOON) (CET) April 1, 2016.  (For time management reasons, we will not correct assignments before this date, so wait with your hand-in until then.) This is a hard deadline for hand-in of assignments for this course round. After this, students will have to wait for the next round of the course to present their assignments. Both assignments have to be presented during the same course round, i.e., single assignments can not be saved between rounds.

The grades on the individual assignments are weighted together as explained in the table below. NOTE: You need to pass both assignments, i.e., get at least grade E on both assignments.