Information theory and coding

COM-404

This file is part of the content downloaded from Information theory and coding.


Course Instructor
  • Emre Telatar, INR 117, emre.telatar@epfl.ch

Teaching Assistants
  • Serhat Emre Coban, INR 036, serhat.coban@epfl.ch
Lectures
  • Monday, 11h15-13h00, BC 03
  • Tuesday, 13h15-15h00, MXG110

Exercise session
  • Tuesday, 15h15-17h00, CM 013

Grading scheme
  • Midterm 40% (October 28th, 2025, 13h15 to 16h15 (location: SG0213))
  • Graded Homework 10% (tentatively due mid-December)
  • Final 50% (To be announced by SAC)

Textbook
  • T. M. Cover and J. A. Thomas, Elements of Information Theory, Wiley.


(8 september)
Source coding / Data compression
(Chapter 5 of the textbook):
- Injective, uniquely decodable, prefix-free (binary) codes
- Kraft sum, Kraft inequalities

(9 september)
- (Partial) converse to Kraft inequality
- Expected codeword length: lower bound


(15 sept) 
- Expected codeword length: lower and upper bounds, asymptotic per-letter tightness
Information measures (Chapter 2 of the textbook):
- Entropy
- KL divergence
- Huffman code/algorithm

(16 sept)
- Some properties of Entropy
- Conditional Entropy, Joint Entropy, Chain rule for entropy
- Source coding with side information
- Mutual information, conditional mutual information
- Properties of mutual information
- Chain Rule for mutual information


(22 sept)
- Public Holiday in Vaud (No lecture)

(23 sept)
- Examples and properties on information theoric measures (entropy, mutual info, etc.)
- Data processing inequality
- Markov Chains

(29 sept)
- Entropy on stationary processes
- Typicality

(30 sept)
- Properties of typical sets/sequences
- Entropy rate
- KL divergence as regret
- minmax regret

(6/10)
- More on entropy rate
- Universal compression Example
- Lempel-Ziv

(7/10)
- Lempel-Ziv
- Finite State Machines (FSM)
- Some related concepts (Information Lossless (IL), Distinct Parsings)
- Analysis of Lempel-Ziv algorithm in comparison to IL-FSM


(13 October)
- Universal compression and prediction (connections to learning theory)
- Data Transmission
- Channels and Capacity

(14 October)
- Stationary, memoryless channels without feedback
- Fano Inequality
- Channel coding with stationary sources
- Relating error-rate with entropy-rate and channel capacity 



[Break] 20 October - 26 October


[Midterm week] 27 October - 2 November

(27 October)
- Converse theorem of channel coding

(28 October) Midterm exam, 13h15 to 16h15, in SG0213 (note the different location)
- You are allowed a single A4 sheet (2 sides) as a cheatsheet. This may be prepared however you like --- hand-written on paper, printed from tablet, LaTeX, and so on --- but you are strongly encouraged to make your own.
- Seven previous years' midterms and their solutions have been uploaded below, for your practice.


(3 November)
- KKT conditions for capacity

(4 November)
- Random coding argument to show achievability of coding theorem

(10 november)
- Channel coding: good news proof

(11 november)
Differential entropy (Chapter 8 of the textbook):
- Definition

(17 november)
- Properties of differential entropy

(18 november)
- Gaussian channel

(24 November)
- Parallel Gaussian channel

(25 November)
Lossy compression (Chapter 10 of the textbook):
- Rate-distortion theory

[Graded HW] 4 December - 10 December

1 December
- Good news theorem of rate-distortion theory

2 December
Rudimentary coding theory (Notes on coding, Moodle---posted in next week's section)

8 December
- Coding theory

9 December
- More coding theory



15 December
- Polar codes

16 December
- Polar codes


[Final month] 1 January - 22 January

Final exam, January 22 Thursday, 9h15 to 12h15, in CM 1106 
- You are allowed two A4 sheets (total of 4 sides) as a cheatsheet. This may be prepared however you like --- hand-written on paper, printed from tablet, LaTeX, and so on --- but you are strongly encouraged to make your own.
- Eight previous years' finals and their solutions have been uploaded below, for your practice.