Assignments

Dates

• See the calendar page for assignment dates.
• All assignments are due on a Tuesday, in class, on paper, at the start of class, i.e. 1:10pm. If that is somehow impossible, contact me to explain the situation, and we'll figure something out.
• Assignments will be posted on this web page, immediately after the Tuesday lecture, almost exactly one week before they are due. They will then be discussed in a tutorial that Thursday. The Thursday after an assignment is due, the tutorial will again be about that assignment.

Marking scheme and lateness penalties

• Each of the 4 assignments will be worth 10% of the final grade.
• Except in the case of an official Student Medical Certificate, assignments that are submitted late will be graded out of 70%, 50%, or 0% of the full score, depending on whether they are 1, 2, or more days late. The time past the deadline will be rounded UP to an integer number of days.

Collaboration Policy for Assignments

• You are expected to work on the assignments by yourself. You should not discuss them with anyone except the tutors or the instructor. The report you hand in should be entirely your own work and you may be asked to demonstrate how you got any results that you report.

What will be in the assignments

• A typical assignment will require you to write (or modify) and use some Matlab code that implements a simple version of a learning procedure that has recently been covered in the course. You will have to submit a very brief report (one page plus figures) that describes the results you obtained.
  
  
• Assignment 1 involves using the backpropagation algorithm to learn distributed representations of words.
  
  
• Assignment 2 focuses on backprop, simple optimization, and simple regularization.
  
  
• Assignment 3 involves learning a mixture of Gaussians.
  
  
• Assignment 4 involves learning a Restricted Boltzmann Machine and using it to improve backpropagation.

CSC321 - Introduction to Neural Networks and Machine Learning