Teaching · Fall 2026

CSCI 8945 — Advanced Representation Learning

Advanced Representation Learning examines how machines learn to represent data — classical and deep methods alike — and how those representations power downstream tasks in vision, language, audio, and beyond. Through a research project, students develop and evaluate new representation-learning ideas and write them up as a manuscript. By the end of the course, students will understand the state of the art in representation learning and be able to apply it to real problems.

At a glance

Meets

Tue & Thu, 1:15–2:35 PM (lecture), Miller Plant Sciences 1503

Wed, 1:15–2:10 PM (discussion), Forest Sciences-4 0516

Instructor

Prof. Jin Sun

Office hours: Tue 3-4pm or by appointment

jinsun@uga.edu

Course info

4 credit hours

Prerequisites: machine learning basics and relevant math (linear algebra, calculus, probability)

Learning outcomes

  1. Demonstrate understanding of machine learning and deep neural network fundamentals.
  2. Gain experience deploying deep learning models in computer vision, natural language processing, and audio domains.

Reading materials

No single required textbook — these are useful free references throughout the semester.

Evaluation and grading

ComponentWeight
Homework (includes individual paper presentation)40%
Midterm exam (in-class)20%
Team project & final presentation40%

Team project

Students work in teams of 3–4 on a research project, ideally applying deep learning and representation learning to a real-world problem. Any programming language or software stack is fine. Wednesday sessions throughout the semester are reserved for project and research discussion — use them for feedback and troubleshooting as the project develops. All teams present their final work in a single in-class session on the last day of class, Thu Nov 19.

Individual paper presentations

Starting mid-October, each class period in the "Advanced topics" unit is themed around a research area (e.g., self-supervised learning, generative models, reasoning). Students individually choose and present one research paper related to that day's theme, leading discussion with the class. With roughly 22 students across 10 themed sessions, expect about 2 presenters per session.

Class schedule

Regular lectures run Aug 18 – Nov 17; final project presentations are Thu Nov 19, the last day of class for this meeting pattern. All dates verified against the UGA academic calendar.

DateDayTypeTopic
Introduction
Aug 18TueLectureIntroduction and overview
Aug 19WedLectureCourse topics
Aug 20ThuLectureData and dimensionality
Aug 25TueLectureDimension reduction, metric learning, PCA, MDS
Aug 26WedDiscussionHow to do research?
Aug 27ThuLectureStructures in data spaces, manifolds, subspaces, sparse coding
Vision
Sep 1TueLecturePixels, 3D points, and cameras
Sep 2WedDiscussionHow to read a paper?
Sep 3ThuLectureImage operations and image semantics
Sep 8TueLectureVideos
Sep 9WedDiscussionHow to write a paper? / project proposal guidance
Sep 10ThuLectureImage subspaces and manipulations
Language
Sep 15TueLectureRepresenting words and sentences
Sep 16WedDiscussionProject and research discussion
Sep 17ThuLectureLanguage model pretraining
Sep 22TueLectureNLP tasks
Sep 23WedDiscussionProject and research discussion
Sep 24ThuLectureZero-shot and in-context learning
Audio
Sep 29TueLectureRepresenting sound
Sep 30WedDiscussionProject and research discussion
Oct 1ThuLectureAudio generation and editing
Multi-modal
Oct 6TueLectureOverview of multi-modal learning
Oct 7WedDiscussionProject and research discussion
Oct 8ThuLectureMultimodal representation alignment
Advanced topics — student paper presentations
Oct 13TueMidtermIn-class midterm exam
Oct 14WedDiscussionProject and research discussion
Oct 15ThuPresentationsUnsupervised learning
Oct 20TuePresentationsSelf-supervised learning
Oct 21WedDiscussionProject and research discussion
Oct 22ThuPresentationsDomain adaptation and transfer learning
Oct 27TuePresentationsLLM finetuning
Oct 28WedDiscussionProject and research discussion
Oct 29ThuPresentationsGenerative models
Nov 3TuePresentationsAI agents
Nov 4WedDiscussionProject and research discussion
Nov 5ThuPresentationsTime series
Nov 10TuePresentationsUnderstanding LLM, VLM, etc.
Nov 11WedDiscussionProject and research discussion
Nov 12ThuPresentationsGraphs / graph representation learning
Nov 17TuePresentationsReasoning
Nov 18WedDiscussionProject and research discussion / final presentation prep
Project
Nov 19ThuPresentationsFinal project presentations (in class, 1:15–2:35 PM)
Nov 23–27Thanksgiving break, no class
Dec 2WedReading day
Dec 3–9Final exam period (no additional meeting for this course)

Thu Nov 19 is this course's last class meeting. Dec 1's university-wide makeup schedule applies only to WF/Friday-only meeting patterns, not Tue/Wed/Thu.