Note: bCourses enrollment is not needed as of now. All materials will be available in shared Google Drive folders (links below) with your @berkeley.edu.
Please, put your name in one of two scribe roles (login with @berkeley.edu). See the information in scribes section.
Please, enroll in ED!
If you find any typos, mistakes or inaccuracies in scribes, please mention them in ED.
Instructor: Nikita Zhivotovskiy
Office hours: Tuesday 2:00 PM - 4:00 PM (315 Evans)
Use ED for questions. Home Assignments are in the Course Materials folder (requires @berkeley.edu account)
No Final Written Exam.
The final project is based on a presentation of a Stat 260 or Stat 210B related paper or a book chapter.
Give a 7-minute focused talk highlighting the key mathematical point or result in the chosen paper. This format mimics a short presentation at large conferences.
Choose a paper either from the provided list (which will be updated periodically) or propose a paper you wish to present. The only conditions are that no two people may work on the same paper, and the paper content should be relevant to the theoretical topics covered in class.
Prepare a 5-7 page, self-contained mathematical text focusing on the core argument. Emphasis should be on clear exposition, simplifying proofs, refining notation, and working through illustrative examples. Feel free to focus on a specific part of the paper that excites you most, such as an interesting special case of a broader result. The goal is to create an expository note that could be valuable to a wide audience. Aim for a quality standard that you would be proud to display on your website as teaching or expository materials.
Timeline:
Final Presentations: During lectures on December 3 and December 5.
Final Document Submission: Due Friday, December 13, serving as the final exam.
This is an advanced, fast-paced graduate course focused on non-asymptotic techniques in high-dimensional statistics. The course complements Stat 210B, but does not require it as a prerequisite. List of topics:
Stein's unbiased risk estimate and its applications.
Minimax lower bounds.
RKHS theory and its relation to statistics.
Sparse recovery.
Elements of sampling theory.
Analysis of interpolating estimators.
Basics of generic chaining and its applications (if time permits).
There will be two homework assignments during the course.
All necessary materials and resources will be available in the shared Google Drive folders.
Assignments must be submitted by 11:59 PM on the due date. Late submissions will not be evaluated, except in cases of emergencies.
Homework must be submitted in digital text format only. Use LaTeX for written assignments.
Students will be responsible for reviewing and grading their own homework using provided solution guides and rubrics.
Regrade Option:
Students have the opportunity to request a regrade for problems they initially solved incorrectly or were unable to solve.
To qualify, students must submit an explanation of their mistakes and their version of the correct solution.
Students get 50% of the score after the resubmission.
Attendance is not mandatory but expected and is strongly encouraged. This course does not offer video recordings.