Coursework

Assignment calendar

Homework materials and submission links are added as assignments are released.

Work Weight Target deadline Materials
Homework 0: mathematical readiness 0% Sun, Sep 6, 11:59 p.m. PDF · LaTeX · Submit on Canvas
Homework 1: prediction, generalization, and evaluation 10% Fri, Sep 18, 11:59 p.m. PDF · LaTeX source bundle · Submit on Canvas
Homework 2: sequence objectives, state, and dependence 10% Fri, Oct 2, 11:59 p.m. PDF · LaTeX source bundle · Submit on Canvas
Research-question first draft Part of 10% Sun, Oct 11, 11:59 p.m. Template · Submit on Canvas
Homework 3: distribution shift, learnability, and interpolation 10% Fri, Oct 16, 11:59 p.m. PDF · LaTeX source bundle · Submit on Canvas
Homework 4: optimization geometry and shared representations 10% Fri, Nov 6, 11:59 p.m. PDF · LaTeX source bundle · Submit on Canvas
Homework 5 10% Fri, Nov 20, 11:59 p.m. To be posted
Research-question revised note Part of 10% Sun, Nov 29, 11:59 p.m. Guidelines · Submit on Canvas
Homework 6 10% Fri, Dec 4, 11:59 p.m. To be posted
Final oral exam 20% Final-examination period Topic list posted at least one week before

The raw weights shown for the six homeworks are each 10 percentage points. The two-lowest-grade replacement described in the syllabus is applied before the homework component is calculated.

Late homework is not accepted. The two-lowest-grade replacement covers ordinary illness, overload, and missed work.

Scribe-note submission

Scribe notes are not submitted through Canvas. Send the completed note to the instructor in a private message on the course Slack. Attach both the LaTeX source (.tex) and the compiled PDF (.pdf). The syllabus gives the deadline, template, and content requirements.

What homework is for

Homework serves three purposes:

  1. Practice: reconstruct definitions and proofs until the ideas become usable.
  2. Diagnosis: discover gaps in prerequisites or understanding while there is time to repair them.
  3. Communication: explain mathematical claims precisely, including assumptions and limits.

A typical two-week assignment will have about four substantial problems, usually with several parts. Some problems will develop one result in stages. Others will ask for a counterexample, a theorem audit, or a short explanation connecting a mathematical model to a current AI system.

Reading an exercise profile

Exercise profiles describe the work before you begin it. Filled mountain pictograms rate difficulty from a direct application to a substantial challenge. Filled cog pictograms rate the amount of formal calculation, proof, or bookkeeping. Filled hourglasses show estimated time: one means under 20 minutes, two means 20–45 minutes, three means 45–90 minutes, and four means more than 90 minutes. Pale pictograms show the unused part of each scale. Work pictograms always appear in this order when applicable: proof, calculation, construction, computation, interpretation. The target pictogram introduces the intended learning outcome. Time estimates cover the non-optional parts. These ratings are not prerequisites or grading weights.

Submission format

Use the homework LaTeX template. Submit each graded homework through Canvas Assignments as a ZIP archive named hwNN-ccid.zip, where NN is the two-digit homework number and ccid is your University of Alberta CCID. For example: hw01-jsmith.zip.

On a resubmission, Canvas may append a suffix such as -1 to the uploaded archive’s filename. This is expected and does not violate the naming convention. Do not submit again merely to remove the suffix. The required filenames and directory name inside the archive must remain unchanged.

The archive must contain a single top-level directory named hwNN-ccid. This directory must contain at least:

  • hwNN-ccid.pdf: the submitted solution;
  • hwNN-ccid.tex: the main LaTeX source.

Include any additional files needed to compile or reproduce the submission, such as figures, code, or data. Do not include generated LaTeX files such as .aux, .log, or .synctex.gz.

Your name and CCID must appear on the first page of the PDF. Identify collaborators and list all sources and tools used. The mathematical writing must be your own. The most recent archive submitted before the deadline will be graded.

Homework 0

Homework 0 is not graded. It tests proof writing, probability, linear algebra, optimization, and comfort with abstraction. Attempt it during the first week without looking for outside solutions. If substantial parts remain inaccessible after reviewing the relevant prerequisites, speak with the instructor promptly and decide whether the course is a good match.

Submit Homework 0 on Canvas by Sunday, September 6, at 11:59 p.m. Use hw00 in the submission naming convention above.

Homework 1

Homework 1 is due Friday, September 18, at 11:59 p.m. It is worth 100 points and 10% of the course grade. The assignment identifies the required parts and the optional, ungraded extensions. A self-contained LaTeX source bundle is also available. Submit Homework 1 on Canvas.


Back to top

CMPUT 654 · Mathematical Foundations of Modern AI Systems · Fall 2026