Dr. Pavel Dvurechensky

I am a member of the research group Stochastic Algorithms and Nonparametric Statistics of the Weierstrass Institute for Applied Analysis and Stochastics. I am also a member of Math+, the Berlin cluster of excellence.

News

• Joint paper with Egor Gladin, Alexander Mielke, and Jia-Jie Zhu on Interaction-Force Transport Gradient Flows has been accepted to NeurIPS 2024. Preprint.
• New preprint: arXiv:2410.20534 A Cournot-Nash Model for a Coupled Hydrogen and Electricity Market, joint with Caroline Geiersbach, Michael Hintermüller, Aswin Kannan, Stefan Kater, Gregor Zöttl. This work is a part of the Math+ project AA4-13: Equilibria for Distributed Multi-Modal Energy Systems under Uncertainty.
• Joint paper with Eduard Gorbunov, Marina Danilova, Innokentiy Shibaev, and Alexander Gasnikov on High Probability Complexity Bounds for Non-Smooth Stochastic Optimization with Heavy-Tailed Noise has been accepted for publication in Journal of Optimization Theory and Applications. Preprint.
• On 27-29 of August I had a honor to give an invited talk at ALGOPT2024 workshop on Algorithmic Optimization: Tools for AI and Data Science dedicated to Professor Yurii Nesterov, celebrating 50 years of his research career in Optimization. My talk was on several recent results on minimization involving self-concordant functions and barriers. Slides.
• New preprint: ArXiV:2408.11022 on improved global performance guarantees of second-order methods in convex minimization, mainly for self-concordant functions, joint with Yurii Nesterov.
• On 21-26 of July I was at International Conference on Machine Learning in Vienna.
• On 26-28 of June I was at Europt 2024 - 21st Conference on Advances in Continuous Optimization giving an invited session talk on Barrier algorithms for non-convex optimization.
• On 20 and 21 of June I was at French-German-Spanish Conference on Optimization 2024 giving an invited session talk on Decentralized Local Stochastic Extra-Gradient for Variational Inequalities.
• New preprint: ArXiV:2405.17075 on Interaction-Force Transport Gradient Flows, joint with Egor Gladin, Alexander Mielke, and Jia-Jie Zhu.
• Two papers have been accepted to ICML 2024:
  • Barrier Algorithms for Constrained Non-Convex Optimization, joint with Mathias Staudigl. In this paper, we extend the framework of our MathProg paper to the setting of optimization over general convex sets from the setting of convex conic constraints. Preprint.
  • High-Probability Convergence for Composite and Distributed Stochastic Minimization and Variational Inequalities with Heavy-Tailed Noise, joint with Eduard Gorbunov, Abdurakhmon Sadiev, Marina Danilova, Samuel Horvath, Gauthier Gidel, Alexander Gasnikov, and Peter Richtarik. This paper continues the line of works on high-probability bounds in different contexts. Preprint.
• On 8-12 of April I was at the Workshop on Nonsmooth Optimization and Applications (NOPTA 2024) in Honor of the 75th Birthday of Boris Mordukhovich. I had a pleasure to give a semi-plenary talk on our joint works with Mathias Staudigl on barrier algorithms for non-convex problems.
• Together with Matthias Liero, Gabriele Steidl, and Jia-Jie Zhu we organized Workshop on Optimal Transport from Theory to Applications. Slides of the talks are available on the Workshop website.
• Joint paper with Jia-Jie Zhu on Mirror-Prox algorithm for optimization involving measure variable has been accepted to AISTATS 2024. Preprint.
• Joint paper with Mathias Staudigl on barrier algorithms for non-convex conic problems has been accepted for publication in Mathematical Programming. DOI: 10.1007/s10107-024-02062-7.
• On 13.12.2023 I successfully finished my Habilitation process at the Humboldt University of Berlin.

Research interests

• First- and second-order algorithms for convex and non-convex large-scale optimization problems
• Randomized algorithms: random coordinate descent, random (zero-order/derivative-free) directional search
• Algorithms for stochastic optimization
• Optimization under inexact information and with adaptivity to unknown smoothness parameters
• Second-order and higher-order (tensor) optimization methods
• Numerical and complexity aspects of Optimal Transport distances and barycenters
• Algorithms for saddle-point problems, Nash games, and variational inequalities
• Distributed optimization (parallel and decentralized)
• Applications to resource allocation, congested traffic modeling, web-page ranking, machine learning, gas networks equilibrium models

My research is connected to the following Mathematical Research Topics of WIAS

• Stochastic Optimization
• Optimal control of partial differential equations and nonlinear optimization
• Statistical inverse problems

Publications

List of publications

Projects

• MATH+-project EF3-8 "Equilibria for Distributed Multi-Modal Energy Systems under Uncertainty"
  Joint project with M. Hintermüller, C. Geiersbach, and A. Kannan, 2023-2025.
• MATH+-project EF3-8 "Analysis of brain signals by Bayesian Optimal Transport"
  Joint project with K.-R. Müuller, S. Nakajima, and V. Spokoiny, 2021-2022.
• MATH+-project EF3-3 "Optimal transport for imaging"
  Joint project with M. Hintermüller, and V. Spokoiny, 2019-2022.

Short CV

2023	Habilitation in mathematics, Humboldt University Berlin, Title: "Large-Scale Numerical Optimization: Inexact Oracle, Primal-Dual Analysis, Ill-Conditioned Problems"
Since 2015	Research fellow, Research Group 6 "Stochastic Algorithms and Nonparametric Statistics", WIAS, Berlin
2014 - 2015	Research assistant, Institute for Information Transmission Problems, Moscow, Russia
2009 - 2015	Junior researcher, Moscow Institute of Physics and Technology, Moscow, Russia
2013	Ph.D., Moscow Institute of Physics and Technology, Moscow, Russia
2010	Master's Diploma, Moscow Institute of Physics and Technology, Moscow, Russia
2008	Bachelor's Diploma, Moscow Institute of Physics and Technology, Moscow, Russia
Extended CV

Teaching

• Modern Algorithmical Optimization
  Higher School of Economics, Moscow, autumn semester 2021, Online
  
• Theory of optimization algorithms for large-scale problems motivated by machine learning applications
  Humboldt University, Berlin, winter semester 2020/2021, Online
  
• Modern Algorithmical Optimization
  Higher School of Economics, Moscow, autumn semester 2020, Online
  
• Recent developments in optimization methods and machine learning applications
  Humboldt University, Berlin, winter semester 2019/2020
  

Contact details

E-mail	Pavel.Dvurechensky-please remove this text-@wias-berlin.de
Phone	+49 (0) 30 20372 465
Fax	+49 (0) 30 20372 316
Address	Room 212, Weierstrass Institute, Mohrenstrasse 39, 10117 Berlin, Germany
Google Scholar	Pavel Dvurechensky
Researchgate profile	Pavel Dvurechensky
Linkedin profile	Pavel Dvurechensky

Last updated 13.09.2024