Ilya Tolstikhin

Ilya Tolstikhin
Feel free to contact me:  ilya[at]tuebingen[dot]mpg[dot]de

Currently I am a postdoc at the Empirical Inference Department of Max Planck Institute for Intelligent Systems, Tübingen, Germany.
I received a diploma (MSc equivalent) in 2010 from Lomonosov Moscow State University and PhD in 2014 from Dorodnicyn
Computing Center of Russian Academy of Sciences
where I worked with Konstantin Vorontsov on statistical learning theory.

My main research fields are statistical learning theory and theory of machine learning. Particularly, I focus on tight data-dependent generalization error and excess risk bounds
in machine learning, which could shed some light on the process of learning, potentially leading to new and more accurate learning algorithms. Also I am interested in tools, used
to achieve these goals (which are extremely interesting and rich fields of research on their own), including concentration of measure inequalities and empirical process theory.

My CV can be found here.

Publications

Preprints 

Conference papers (chronologically ordered)

Journal papers  (chronologically ordered)

Others 

PhD Thesis


Talks

Talks in English

Talks in Russian


Activities

Journals review: Conferences review: PC Member:

Teaching

Instructor for the course “Machine Learning Theory”

(together with Ruth Urner)

Eberhard Karls Universität Tübingen

2016 - 2017

Teaching Assistant for the course “Machine Learning”

Skolkovo Institute of Science and Technology

2013 - 2013

Tutorials for the course “Machine Learning”

Lomonosov Moscow State University

2012 - 2013

Tutorials for the course “Machine Learning”

Moscow Institute of Physics and Technology

2011 - 2012


Things I'm doing outside of office

Workouts and health tracking: I workout regularly and collect all kind of data I can get about my body and health. In future I hope to run some algorithms on this data to see
if I can build a reasonable model of fitness / health. All the info is available in this text file (except hurt rate listings), which is regularly updated. Feel free to use it ;)