Topics in Statistics (359-0-20)

Topic

Deep Learning

Instructors

Bradly Carson Stadie

Meeting Info

555 Clark B01: Tues, Thurs 12:30PM - 1:50PM

Overview of class

Topic: Modern deep learning.
This class will cover the most essential tools required to use modern deep learning frameworks and read modern deep learning papers. In particular, we will focus on Transformers and Diffusion Models, which are the main foundational pieces of most modern deep learning papers. We will derive both methods from first principles, assuming no prior knowledge of deep learning. We will then investigate various applications of both methods, including large language models, GPT, and stable diffusion. Students will learn to use modern deep learning frameworks including HuggingFace and Collab to run these models at scale.

Registration Requirements

Linear algebra

Learning Objectives

Students should be able to read many modern papers in deep learning by the end of the class. In particular, papers that focus on large language models or diffusion should now be accessible. Further, students will be able to deploy these models using cloud computing resources. Students will also learn how to fine tune these models for their own use case.

Teaching Method

Lecture, Student presentations.

Evaluation Method

Homework, Presentations.

Class Materials (Required)

None required. Papers will be provided.

Class Attributes

Formal Studies Distro Area