Statistical learning theory 2024/25

Материал из Wiki - Факультет компьютерных наук
Версия от 16:41, 18 декабря 2024; Bauwens (обсуждение | вклад)

(разн.) ← Предыдущая | Текущая версия (разн.) | Следующая → (разн.)
Перейти к: навигация, поиск

General Information

Lectures: on Tuesday 9h30--10h50 in room M302 and in zoom by Bruno Bauwens

Seminars: online in Zoom by Nikita Lukianenko.

Please join the telegram group The course is similar to last year.

Results

Problems exam

Friday 20 December 11h-14h, D507
-- You may use handwritten notes, lecture materials from this wiki (either printed or through your PC), Mohri's book
-- You may not search on the internet or interact with other humans (e.g. by phone, forums, etc)

About questions
-- 4 questions of the difficulty of the homework. (Many homework questions were from former exams.)
-- I always ask to calculate VC dimension and to give/prove some risk bound with Rademacher complexity.

Homeworks

Deadline every 2 weeks, before the lecture. The tasks are at the end of each problem list. (Problem lists will be updated, check the year.)

Before 3rd lecture, submit HW from problem lists 1 and 2. Before 5th lecture, from lists 3 and 4. Etc.

Classroom to submit homeworks. You may submit in English or Russian, as latex or as pictures. Results are here.

Late policy: 1 homework can be submitted at most 24 late without explanations.

Course materials

Video Summary Slides Lecture notes Problem list Solutions
Part 1. Online learning
21 Sep Philosophy. The online mistake bound model. The halving and weighted majority algorithms. sl01 ch00 ch01 prob01 sol01
24 Sep The standard optimal algorithm. The perceptron algorithm. sl02 ch02 ch03 prob02 sol02
01 Oct Kernel perceptron algorithm. Prediction with expert advice. Recap probability theory (seminar). sl03 ch04 ch05 prob03 sol03
Part 2. Distribution independent risk bounds
08 Oct Necessity of a hypothesis class. Sample complexity in the realizable setting, examples: threshold functions and finite classes. sl04 ch06 prob04 update 12.10 sol04
15 Oct Growth functions, VC-dimension and the characterization of sample comlexity with VC-dimensions sl05 ch07 ch08 prob05
22 Oct Risk decomposition and the fundamental theorem of statistical learning theory (previous recording covers more) sl06 ch09 prob06 sol06
05 Nov Bounded differences inequality, Rademacher complexity, symmetrization, contraction lemma. sl07 ch10 ch11 prob07 sol07
Part 3. Margin risk bounds with applications
12 Nov Simple regression, support vector machines, margin risk bounds, and neural nets with dropout regularization sl08 ch12 ch13 prob08 sol08
19 Nov Kernels: RKHS, representer theorem, risk bounds sl09 ch14 prob09 sol09
26 Nov AdaBoost and the margin hypothesis sl10 ch15 prob10 sol10
03 Dec Losses of neural nets are not locally convex. Gradient descent with stable gradients. (Old recording about Hessians) ch16 prob11 sol11
10 Dec Lazy training and the neural tangent kernel. ch17
17 Dec Colloquium 9h30 - 12h30 (room D725) and 18h10 - 21h (different building Старая Басманная А-125). Rules and questions. Reserve in shedule.


The lectures in October and November are based on the book: Foundations of machine learning 2nd ed, Mehryar Mohri, Afshin Rostamizadeh, and Ameet Talwalker, 2018.

A gentle introduction to the materials of the first 3 lectures and an overview of probability theory, can be found in chapters 1-6 and 11-12 of the following book: Sanjeev Kulkarni and Gilbert Harman: An Elementary Introduction to Statistical Learning Theory, 2012.

Grading formula

Final grade = 0.35 * [score of homeworks] + 0.35 * [score of colloquium] + 0.3 * [score on the exam] + bonus from quizzes.

All homework questions have the same weight. Each solved extra homework task increases the score of the final exam by 1 point. At the end of the lectures there is a short quiz in which you may earn 0.1 bonus points on the final non-rounded grade.

There is no rounding except for transforming the final grade to the official grade. Arithmetic rounding is used.

Autogrades: if you only need 6/10 on the exam to have the maximal 10/10 for the course, this will be given automatically. This may happen because of extra homework questions and bonuses from quizzes.

Colloquium

Rules and questions from last year.

Date: TBA

Office hours

Bruno Bauwens: Bruno Bauwens: Tuesday 12h -- 20h. Friday 15h -- 17h30. Better send me an email in advance.

Nikita Lukianenko: Write in Telegram, the time is flexible