CS 462: Introduction to Deep Learning

Spring 2026.

Description

As stated on the department website:

This is an introductory course to deep learning. The course will cover theories, principles, and practices of traditional neural networks and modern deep learning. The topics of the course can be split into four sections: (1) Fundamentals of Machine Learning, (2) Neural Networks, (3) Modern Deep Learning, and (4) Applications and Advanced Topics.

Details

Course material will be posted on the course website.

The course text book is “Understanding Deep Learning” by Simon J. D. Prince (udlbook.com).

This course will cover a large swathe of deep learning, corresponding to the first 15 chapters of the book (although topics may be adjusted to better fit the students’ preparedness for the topics). The list of topics is, in a rough summary, the principles of training deep networks (chapters 1-9), convolutional networks (chapter 10), residual networks (chapter 11), transformers (chapter 12), graph neural networks (chapter 13), unsupervised learning (chapter 14), and generative adversarial networks (chapter 15). The list of topics are somewhat aspirational; if more preparation is needed to better cover some fundamentals then not every topic will be reached.

Course Overview

Deep learning is an incredibly approachable field, with tools, datasets, and tutorials widely available. This tricks many people into believing that it is an easy topic to grasp and put into practice. It is not. The pitfalls are plentiful, and mistakes are often made in the very first steps.

This course aims to prepare students to put deep learning into practice by teaching the fundamentals of deep learning. Truly practicing deep learning requires millions of training samples and weeks of time, so an extensive grasp of the topic is beyond reach of this course. This course is an introduction, hence the word appears in the name. However, a proper grasp of the fundamentals is required to advance in this field. This course will prepare students to apply deep learning and will forewarn them of both weaknesses and solutions.

Evaluation

Implementation is an important component of the course, so 40% of the course grade will be based upon individual assignments. Recitation work is meant to prepare students for assignments and examinations, and may consist of shorter quizzes and in-recitation work.

• 30% Final
• 20% Midterm
• 10% Recitation Assignments
• 40% Homeworks

The homeworks are meant to solidify understanding of key concepts and are not intended to be excessively complicated. Some of them will quite simple, so don’t be intimidated by the number of assignments.

Quizzes will be mostly multiple choice questions. Their primary purpose is to give feedback on your current mastery rather than to finalize your score in the class. Full credit will be awarded at 50% and above, and 50% credit will be awarded if any questions are correct.