Deep Learning and Applications Fall 2026

Final project

Team size: 3–4 students
Course grade: 25%
Milestone dates: To be announced

Choosing a project

Choose any topic related to deep learning and investigate a substantive problem beyond the scope of a homework assignment. Your project may take one of the following directions.

Applications: Develop and evaluate a system for a meaningful task.
Methods and systems: Explore changes to a model, training method, or implementation.
Empirical studies: Investigate model behaviour through controlled experiments.

Running existing code alone is not sufficient.

Example: speech enhancement

L01 discusses a student project from Stanford CS230: prepare speech recordings with realistic background noise, train a model and investigate its design choices, then evaluate both noise removal and preservation of the speaker’s voice.

Milestones & grading

These percentages are shares of the final project grade.

Proposal: 20%
Midterm report: 30%
Final report and presentation: 50%

Reports should use a conference-paper format. Submission details will be announced.

What we assess

Substantial work and clear individual contributions.
Understanding of relevant prior work and where your project fits.
A worthwhile, well-defined goal or research question.
Sound experiments and evaluation.
A clear, carefully formatted report.

Teamwork & resources

Find collaborators early. The teaching team can help form teams.
Agree on responsibilities and review the work together.
Keep records of communication and code contributions.
Contact the teaching assistants early if collaboration problems arise.

Computing resources

University cluster access is available for queued GPU jobs.

External materials & AI use

Acknowledge permitted external material and submit work that you understand and wrote. AI tools may be used to explain concepts and clarify lecture material. Do not submit AI-generated answers, derivations, code, or reports.

See L01, slides 87–93, for the project introduction.