Fall 2026: Random Matrix Theory in Data Science and Statistics (EN.553.796)

Course Description

This is a first course in random matrix theory, the study of the eigenvalues and eigenvectors of matrices with random entries that is foundational to high-dimensional statistics and data science. Aside from the main ideas and modern applications of random matrices, a key goal will be to introduce you to the main concepts of probability in high dimensions: concentration of measure, the geometry of high-dimensional spaces and convex sets, Gaussian measure, and sharp transitions and threshold phenomena. The following is a (very) tentative ordered list of specific topics to be covered:

1. Gaussian matrices and dimensionality reduction

  • Coarse properties of rectangular Gaussian random matrices
  • Geometric method of analysis with concentration inequalities
  • Random projections and their dimensionality reduction applications in machine learning

2. Classical theory of i.i.d. random matrices

  • Moment method for eigenvalue limit theorems
  • Semicircle law for Wigner matrices
  • Marchenko-Pastur law for Wishart matrices
  • Elements of universality
  • Elements of free probability theory
  • Applications to neural networks and random optimization landscapes

3. Spiked matrix models

  • Resolvent method for eigenvalue limit theorems and further applications to eigenvectors
  • Baik—Ben Arous—Péché (BBP) phase transition in spiked matrix models
  • Applications in principal component analysis
  • Applications in analysis of neural network weight matrices

4. Matrix concentration inequalities

  • General-purpose concentration inequalities (martingale methods, Lipschitz concentration) and applications to random matrix statistics
  • General-purpose bounds on expected random matrix norms (matrix Chernoff and Bernstein)
  • Non-commutative Khintchine inequality and its recent extensions

Prerequisites: Linear algebra and either Real Analysis (AS.110.405 or equivalent) or Probability Theory I (EN.553.720 or equivalent). You should be familiar with eigenvalues, eigenvectors, and singular value decompositions; formal definitions of limits, convergence, and open and closed sets; laws of large numbers and central limit theorems; and conditional expectation. Deep background in measure-theoretic probability is not necessary, but some topics may be difficult if you are completely unfamiliar with it.

Contact & Office Hours

I am the instructor of this course, Tim Kunisky, and the teaching assistant is AMS PhD student Zhouhao Yang.

The best way to contact us is by email, at kunisky [at] jhu.edu and zyang145 [at] jhu.edu, respectively. We will choose regular office-hour times after the first week of class. In the meantime, please contact us directly to schedule an appointment.

Schedule

Class meets Tuesdays and Thursdays, 9:00am to 10:15am in Maryland 104.

Below is a tentative schedule, to be updated as the semester progresses.

Date Details
Week 1
Sep 1 1. Course logistics. Random vector theory. Concentration inequalities and high-dimensional geometry. Properties of Gaussian random vectors.
Sep 3 2. Orthogonal invariance, Gaussian random vectors, and uniform measure on the sphere. Multiplication by Gaussian random matrices. Johnson-Lindenstrauss lemma and dimensionality reduction. "First moment method" proof technique.
Week 2
Sep 8 3. Finish proof and discussion of Johnson-Lindenstrauss lemma. Extensions and sketch of applications in machine learning. Introduction to "geometric method" for singular values of rectangular matrices.
Sep 10 4. Concentration of singular values of short fat matrices. Interpretation as a "matrix concentration" inequality. Discretizing matrix norms with epsilon nets. Non-constructive proof of existence of good epsilon nets.
Lecture Notes and Materials

You do not need to buy any books for this course. I will post updated lecture notes here as the course proceeds. In the meantime, you can consult the Fall 2025 lecture notes and the Fall 2024 lecture notes, but the topics and presentation may differ this year. If you notice typos in any of the notes, please let me know.

The following are books or lecture notes that cover some similar material and might be useful to you in addition to my notes. These are useful references on general advanced probability; the first is especially friendly:

These are additional references on random matrix theory specifically:

Grading

Grades will be based on a small number of written homework assignments (25%), a written in-class midterm exam (25%), and a final project proposal (5%), report (10%), and presentation (35%) concerning a recent research paper, open problem, or topic of interest related to the material we cover.

The midterm exam will not ask you to solve new problems on the spot. Instead, you will be asked to write up solutions to a selected previous homework problem and to outline the proof to a selected major theorem we have discussed in class.

Policies on Assignments and Collaboration

The following are the policies for submitted work in this course:

  • Collaboration: You are welcome to discuss homework with your classmates and instructors, but you must write up your own solutions, alone, in your own words. Students found submitting verbatim identical solutions will be penalized. If you have discussed the homework with anybody other than instructors, please list their names at the top of your submission.
  • Sharing solutions: You may not share full solutions to homework problems with your classmates, show others your written solutions and ask if they are correct, or take notes on or pictures of other students' work before preparing your own solutions.
  • AI assistants: You are welcome to use AI assistants (ChatGPT, Gemini, Claude, etc.) to explore the topics discussed in lecture, to clarify any general points of confusion, and to ask broad clarifying questions while doing your homework. You are not allowed to ask them directly for the solutions to homework problems. If you discuss your homework with an AI assistant, describe your interaction at the top of your submission. You are allowed to interact with AI assistants on your homework in the same way you are allowed to with other students: you may discuss problems, but may not ask for or copy complete solutions. If in doubt, you should make sure that you would be able to explain your solution to a homework problem on the board with no other references available.
  • Late submissions: You may use a total of five late days for homework submissions over the course of the semester without penalty. If you need an extension beyond these, you must ask me 48 hours before the due date of the homework and have an excellent reason. After you have used up these late days, further late assignments will be penalized by 20% per day they are late (that is, your maximum score after one late day will be 80%, after two late days 60%, and so forth). The final project must be submitted on time and no extensions for it are allowed.
Assignments

Homework will be posted here and submitted through Gradescope (see Canvas announcements for details). Please talk to me in advance if you need more time for an assignment.

Assigned Due Link
Assignment dates and links to be added.
Final Project

Your final project is to do one of the following on a topic related to the content of this course: (1) read and digest a paper and present its content in your own words and style, with some elaboration that is not present in the original paper; (2) perform an interesting computational experiment motivated by something we have seen in class or something you read in a paper and report in detail on the results and their interpretation; or (3) for the intrepid, find an open problem related to something we have seen in class, try to work on it, and report your findings.

The project submission will have three parts:

  • An initial proposal of 0.5-1 pages describing what you plan to do.
  • A short blackboard presentation during the assigned final exam period.
  • A short written report of 4-6 pages.

More detailed requirements, due date for the proposal, and suggestions of reasonable project topics will be posted later.