Statistics 311/Electrical Engineering 377: Information Theory and Statistics

John Duchi, Stanford University, Fall 2025

Approximate Course Schedule

The syllabus below suggests what will (likely) be our approximate course schedule. We will likely change a few things around as the course continues, and we may even omit topics or add others as the class desires. While we skip some chapters, we encourage students to at least skim through them (for example, Chapter 3 on exponential families will provide useful background, especially if students have not seen them before). When reading is optional (but provides good context), we will add an asterisk (*) to it.

Lecture Date Topics Reading
1 Tue, Sep 22 Overview, basic divergence measures LN 1-2, CT 2*
2 Thu, Sep 24 Chain rules and general divergence measures LN 2, CT 2
3 Tue, Sep 30 Le Cam and Fano inequalities, concentration LN 2, 4.1, RM 2.3
4 Thu, Oct 2 Sub-exponential concentration LN 4.1, 4.2, RM 2.4
5 Tue, Oct 7 Martingale methods and uniformity LN 4.2, 4.3, RM 2
6 Thu, Oct 9 Uniform laws, beginning PAC-Bayes bounds LN 5.1, 5.2
7 Tue, Oct 14 PAC-Bayes bounds and bits of interactive data analysis LN 5.2
8 Thu, Oct 16 Interactive data analysis LN 5.3
9 Tue, Oct 21 Privacy and disclusure limitation LN 7.1, 7.2
10 Thu, Oct 23 Privacy: composition guarantees LN 7.2, 7.3
11 Tue, Oct 28 Privacy: inverse sensitivity LN 7.4
12 Thu, Oct 30 Le Cam/Fano methods LN 8.1–8.4
Tue, Nov 4 No class (democracy day)
13 Thu, Nov 6 Assouad's method LN 8.5–8.6
14 Tue, Nov 11 Strong data processing inequalities LN 9.1–9.2
15 Thu, Nov 13 Constrained lower bounds LN 9.1–9.2
16 Tue, Nov 18 Loss functions and entropy LN 11.1–11.3
17 Thu, Nov 20 Calibration and proper losses LN 12.1–12.3
18 Tue, Dec 2 Surrogate risk consistency
19 Thu, Dec 4 Presentations

Abbreviation key