EC5203 · EC4310 · MM4450 · Fall 2026 · Heung-No Lee

Information Theory

Entropy, data compression, and channel capacity — from Shannon's 1948 framework to its role in modern AI. 26 lectures, Mon & Wed, 대학 C동 110호.

C = maxp(x) I(X;Y)

Aug 31 – Dec 18, 2026 Cover & Thomas 2 Quizzes · Midterm · Final

Weekly Schedule

Same routine as 2025: 14 lectures before the midterm, 12 after. Lecture notes and full videos on the Heung-No Lee Lab YouTube channel. Dates follow the official GIST 2026 academic calendar: semester Aug 31 – Dec 18, midterm week Oct 19–23, final week Dec 14–18. Chuseok (Sep 24–26, Thu–Sat) does not affect Mon/Wed classes; Oct 5 (Mon) is the substitute holiday for National Foundation Day — study the assigned videos that week.

🗂 Course Materials — Lecture Slides, Lecture Book, and all shared class files are in the class Google Drive. Open Drive →
WkMondayWednesdayAssignments due Wed 23:59
18/31
Introduction to class
9/2
Probability & Random Variables
29/7
LLM · Next-Token Prediction
9/9
Information / Horse Race
HW0 not graded
39/14
Entropy
9/16
Mutual Information
HW1
49/21
Jensen's Inequality · Relative Entropy
9/23
DPI · Markov Chains · 2nd Law
추석연휴 9/24–26 (목–토)
HW2
59/28
Sufficient Statistics · Fano · Transformers
9/30
Quiz #1 · Convergences · LLN
Quiz 1
610/5
대체휴일 (개천절) — no class.
Videos: AEP · Typical Sets
10/7
Entropy Rate · Cesàro Mean
HW3
710/12
Data Compression · Kraft Inequality
10/14
Huffman · Lempel-Ziv · Midterm Review
HW4
810/19 – 10/23
Midterm Exam (GIST midterm week)
910/26
Communications over Channels
10/28
Hamming Codes · Erasure Correction
1011/2
Channel Capacity · Noisy Typewriter
11/4
BSC · BEC · Symmetric Channels
HW5
1111/9
Channel Coding Theorem: Setup & Key Ideas
11/11
Forward Proof · Information BER
HW6
1211/16
Converse Proof (Fano's Inequality)
11/18
Quiz #2 · Differential Entropy
Quiz 2
1311/23
Gaussian Channel Motivation
11/25
AWGN Channel Capacity
1411/30
Gaussian Coding Theorem · Sphere Packing
12/2
Parallel Gaussian Channels · Water-filling
HW7
1512/7
Multiple Access · Slepian-Wolf
12/9
Final Preparation
HW8
1612/14 – 12/18
Final Exam (GIST final week)

Homework Submission via GitHub

Homework is due 23:59 KST on Wednesday — the end of Wednesday, not its beginning. Grading collects your repository at that moment, so anything pushed later is not graded. You register your GitHub username once and use one private repository all semester; the AI grader commits corrections there and opens an issue that mentions you.

1 · Register

  1. Create a GitHub account if you do not have one. Any username is fine.
  2. Write your username next to your student ID in the registration sheet by Monday of week 2 — no username means no repository, and no way to submit.
  3. Accept the emailed invitation to it2026-<your-username>.

2 · Submit

  1. All ten problem sheets are already in problems/ in your repository.
  2. Put your solution in submission/hwN.pdf — exactly that name, N being the assignment number. A file named anything else is not found, and counts as missing.
  3. PDF (scanned handwriting is fine if legible) or Markdown. Deadlines are always Wednesday 23:59 KST; the last commit before that moment is the one graded.

3 · Feedback

  1. Per-problem scores and corrections appear as feedback/hwN-feedback.md.
  2. An issue titled HW#N graded: score/100 is opened and mentions you.
  3. Regrade requests: reply in that issue within 3 days — your TA is notified there, so there is no one to email.

AI Teaching Assistant

Two channels. Quick questions — definitions, theorem statements, course logistics — get instant answers right here in this chat. For rigorous full proofs, deep dives, or homework disputes, use the GitHub Q&A (open an issue and mention @claude): a top-tier model answers there, and every Q&A stays public so you can search past answers. The TA gives hints, not homework solutions.

AI TA
Hello — I am the AI TA for Information Theory 2026. Ask me anything about the lectures, Cover & Thomas, or course logistics. For long rigorous proofs, I may point you to the GitHub Q&A, where a stronger model answers in full.

Course Grades

Midterm 30
Final 30
HW + Reviews 20
Att. 10
Quiz 10

Attendance: no-show without prior notice, or late entry into (Zoom) class, is considered not polite.

Resources

Materials

Contact

  • Prof. Heung-No Lee, EECS, GIST
  • Office hours: after class or by appointment
  • Homework disputes: GitHub Issue in your repo