Simon Fraser University
Engaging the World

Course Outlines

Fall 2026 - CMPT 419 D500

Special Topics in Artificial Intelligence (3)

Trustworthy Deep Learning

Class Number: 6560

Delivery Method: In Person

Overview

  • Course Times + Location:

    Sep 9 – Sep 15, 2026: Tue, 2:30–5:20 p.m.
    Burnaby

    Sep 22 – Dec 6, 2026: Tue, 2:30–5:20 p.m.
    Burnaby

Description

CALENDAR DESCRIPTION:

Current topics in artificial intelligence depending on faculty and student interest.

COURSE DETAILS:

Topic: Trustworthy Deep Learning
Prerequisite: CMPT 225

This is a seminar-style special topics course on trustworthy AI and risk and robustness of LLMs. The purpose is to promote the awareness of several important issues in using AI technology  and LLMs, including safety/security, fairness, data privacy, and to understand technical solutions to these problems. Teaching materials include a collecction of selected survey papers and research papers on these topics. Working knowledge of machine learning techniques is assumed.

The course has three components: (1) the instructor will first present the key concepts/techniques of trustworthy AI in the first half of the course; (2) the student will present papers chosen from a recommended list in  the second half of the course, each student has two presentations; (3) each student will conduct a research project on a problem related to trustworth AI.  Required and recommended reading lists, and lecture notes will be available at sfu.canvas/Files at the beginning of Fall 2026.

Topics:

Overview on trustworthy AI
Attacks and defenses for machine learning models
Attacks on LLMs
Safety alignment of LLMs
Attacks and defenses for multimodal LLMs
Fairness of machine learning models
Data privacy for machine learning models

COURSE-LEVEL EDUCATIONAL GOALS:

To promote responsible design and use of AI technology and LLMs.

Grading

  • Class participation 20%
  • Research paper presentation 20%
  • Project proposal 10%
  • Project presentation 20%
  • Project report 30%

Materials

REQUIRED READING:

Selected survey and research papers publicly available. The list will be provided at the beginning of the course. 

Department Undergraduate Notes:

The following are default policies in the School of Computing Science. Please check your course syllabus whether the instructor has chosen a different policy for your class, otherwise the following policies apply.

  • Students must attain an overall passing grade on the weighted average of exams in the course in order to get a D or higher.

 

  • All student requests for accommodations for their religious practices must be made in writing by the end of the first week of classes, or no later than one week after a student adds a course. After considering a request, an instructor may provide a concession or may decline to do so. Students requiring accommodations as a result of a disability can contact the Centre for Accessible Learning (caladmin@sfu.ca).

 

  • Use of AI tools by students for completing graded scholarly activities and course submissions that has not been explicitly authorized by the instructor  will be considered as plagiarism under SFU’s Code of Academic Policy (http://www.sfu.ca/policies/teaching/t10-02.htm).  Following this policy, instructors will report all instances of plagiarism to the academic integrity board.  

Registrar Notes:

ACADEMIC INTEGRITY: YOUR WORK, YOUR SUCCESS

At SFU, you are expected to act honestly and responsibly in all your academic work. Cheating, plagiarism, or any other form of academic dishonesty harms your own learning, undermines the efforts of your classmates who pursue their studies honestly, and goes against the core values of the university.

To learn more about the academic disciplinary process and relevant academic supports, visit: 


RELIGIOUS ACCOMMODATION

Students with a faith background who may need accommodations during the term are encouraged to assess their needs as soon as possible and review the Multifaith religious accommodations website. The page outlines ways they begin working toward an accommodation and ensure solutions can be reached in a timely fashion.