Consultation hours

Friday
From 14:00 to 18:00
Office S 943 on Pokrovsky campus 
For another appointment, contact me by email at qparis@hse.ru
 

Current courses

Topics in High-Dimensional Probability and Statistics
​
Main reference:
R. Vershynin. High-dimensional probability. An introduction with applications in data science. Cambridge Series in Statistical and Probabilistic Mathematics. vol 47. 2018 
​
Journal:
 
Lecture 8:
High-dimensional linear regression; Lasso procedure.
Black board: [pdf]
Notes: [pdf]
​
Lecture 7:
Stochastic bloc model; Community detection in random graphs.
Notes: [pdf]
​
Lecture 6:
Matrix Bernstein inequality.
​
Lecture 5:
PCA; Spiked covariance model.
​
Lecture 4:
Johnson-Lindenstrauss lemma. Covariance matrix estimation.
Notes: [pdf]
​
Lecture 3:
Suprema of sub-gaussian random variables.
Notes: [pdf]
​
Lecture 2:
Generalized Hoeffding inequality for sums of independent sub-gaussian r.v.'s; Sub-exponential distributions; Bernstein inequality for sums of independent sub-exponential r.v.'s.
Notes: [pdf]
​
Lecture 1:
Elementary properties of sub-gaussian distributions and in particular their concentration properties.
Notes: [pdf]
​
​
​
Topics in Statistical Learning Theory
(Modern Methods of Decision Making)
​
Main reference:
These lectures: (link)
Older lecture notes: (link)
​
Journal:
​
Lecture 9: 
Mirror descent.

Time: Monday, April 6, 2020, 13:30 (Moscow)

Join Zoom Meeting: here

Meeting ID: 807 937 894

Password: 133598

​
Lecture 8: 
Gradient descent for smooth and strongly convex functions; Accelerated gradient descent.
Notes: [pdf]
​
Lecture 7:
Gradient descent for convex functions.
Notes: [pdf]
​
Lecture 6:
Introduction to convex optimization.
Notes: [pdf]
​
Lecture 5:
Convex approach to binary classification.
​
Lecture 4:
Rademacher complexity and VC dimension.
​
Lecture 3:
Introduction to empirical risk minimization (ERM); Estimation-Approximation tradeoff; ERM with a finite class and a bounded loss; Noiseless case.
​
Lecture 2:
Conditional probabilities and expectation; Optimal predictors; Examples of the square and binary losses.
​
Lecture 1:
Introduction to supervised learning; Learning sample; Loss functions, Risk and Excess risk.