Elective400 LevelSecond SemesterArtificial Intelligence
CSC404
Deep Learning & Neural Networks
Course Description
Deep Learning & Neural Networks goes past the introductory treatment in CSC304 into architectures for images, sequences, and beyond, along with the practical craft of actually training these models: initialization, regularization, and diagnosing why a network refuses to learn.
Learning Outcomes
- Explain the architecture and training process of feedforward and convolutional networks.
- Explain architectures for sequential data and their common failure modes.
- Diagnose common training problems and apply appropriate fixes.
- Critically evaluate a deep learning result reported in a paper or product claim.
Weekly Topics
Examines neural network foundations review from a practical angle, using a case drawn from real systems to motivate the concepts.
Builds directly on the previous week to extend backpropagation in depth, with an emphasis on where the earlier techniques stop working.
Combines a short lecture on convolutional networks with an in-class exercise students carry into the week's assignment.
Focuses on common mistakes and misconceptions around training dynamics & optimization, using student work from the previous assignment as material.
Introduces regularization techniques and immediately puts it to use in a small design or implementation task.
Situates sequential models within the broader arc of the course, showing how it connects to what comes next.
Uses a guest dataset or scenario to explore mid-semester review in a setting closer to professional practice.
Introduces attention mechanisms through short lectures and worked examples, building the vocabulary the rest of the course relies on.
Works through transfer learning in a lab-driven session, with guided exercises students complete and discuss in small groups.
Covers model evaluation & failure analysis in depth, connecting the underlying theory to a concrete example the class builds together.
Examines applied project i from a practical angle, using a case drawn from real systems to motivate the concepts.
Builds directly on the previous week to extend applied project ii, with an emphasis on where the earlier techniques stop working.
Combines a short lecture on revision with an in-class exercise students carry into the week's assignment.
Assessment
Assignments — 20%
Practical Work — 35%
Mid-Semester — 10%
Final Examination — 35%
Prerequisite Map
Related Courses