Talks

[29] High-Probability Convergence for Composite and Distributed Stochastic Minimization and Variational Inequalities with Heavy-Tailed Noise
ICML 2024 Oral, Vienna, Austria, 25 July, 2024
[slides]

[28] Byzantine Robustness and Partial Participation Can Be Achieved Simultaneously: Just Clip Gradient Differences
EURO 2024, Copenhagen, Denmark, 3 July, 2024
[slides]

[27] Last-Iterate Convergence of Extragradient-Based Methods
EUROPT 2024, Lund, Sweden 26 June, 2024
[slides]

[26] Byzantine Robustness and Partial Participation Can Be Achieved Simultaneously: Just Clip Gradient Differences
Invited talk at INSAIT, 24 June, 2024
[slides]

[25] Byzantine Robustness and Partial Participation Can Be Achieved Simultaneously: Just Clip Gradient Differences
PODL 2024, Nantes, France, 21 June, 2024
[slides]

[24] Byzantine Robustness and Partial Participation Can Be Achieved Simultaneously: Just Clip Gradient Differences
NETYS 2024, online, 29 May, 2024
(Keynote talk)
[slides]

[23] Variance Reduction for Byzantine-Robust Distributed Optimization
Federated Learning One-World Seminar, online, 7 February, 2024
[slides] [video]

[22] "Clipped Methods for Stochastic Optimization with Heavy-Tailed Noise"
TES Conference on Mathematical Optimization for Machine Learning, Berlin, Germany, 15 September, 2023
[slides]

[21] "Algorithms for Stochastic Optimization with Heavy-Tailed Noise and Connections with the Training of Large Language Models"
Oberseminar at LT Group, University of Hamburg, Germany, 6 June, 2023
[slides]

[20] "Convergence of Proximal Point and Extragradient-Based Methods Beyond Monotonicity: the Case of Negative Comonotonicity"
PEP talks, UCLouvain, Belgium, 13 February, 2023  
[slides] [video]

[19] "Variance Reduction is an Antidote to Byzantines: Better Rates, Weaker Assumptions and Communication Compression as a Cherry on the Top"
MBZUAI Workshop on Collaborative Learning: From Theory to Practice, Abu Dhabi, UAE, 8 October, 2022 
[slides]

[18] "Methods with Clipping for Stochastic Optimization and Variational Inequalities with Heavy-Tailed Noise"
All-Russian Optimization Seminar, online, 9 September, 2022 (in Russian)
[slides] [video]

[17] "Distributed Methods with Absolute Compression and Error Compensation"
MOTOR 2022, Petrozavodsk, Russia, 3 July, 2022
[slides]

[16] "Secure Distributed Training at Scale"
Lagrange Workshop on Federated Learning, online, 25 April, 2022 
[slides]

[15] "Extragradient Method: O(1/K) Last-Iterate Convergence for Monotone Variational Inequalities
and Connections With Cocoercivity"
Rising Stars in AI Symposium 2022, KAUST, Saudi Arabia, 13 March, 2022
[slides] [video]

[14] "Moshpit SGD: Communication-Efficient Decentralized Training on Heterogeneous Unreliable Devices"
Vector Institute Endless Summer School session "NeurIPS 2021 Highlights", online, 16 February, 2022 
[slides] 

[13] "Moshpit SGD: Communication-Efficient Decentralized Training on Heterogeneous Unreliable Devices"  
MLO EPFL internal seminar, online, 20 December, 2021
[slides]

[12] "Extragradient Method: O(1/K) Last-Iterate Convergence for Monotone Variational Inequalities
and Connections With Cocoercivity"
MTL MLOpt internal seminar, online, 1 December, 2021 
[slides]

[11] "Extragradient Method: O(1/K) Last-Iterate Convergence for Monotone Variational Inequalities
and Connections With Cocoercivity"
All-Russian Optimization Seminar, online, 17 November, 2021 (in Russian)
[slides] [video]

[10] "Secure Distributed Training at Scale"
Federated Learning One-World Seminar, online, 3 November, 2021
[slides] [video]

[9] "MARINA: Faster Non-Convex Distributed Learning with Compression"
Federated Learning One-World Seminar, online, 10 March, 2021 
[slides] [video]

[8] "Linearly Converging Error Compensated SGD"
NeurIPS New Year AfterParty at Yandex, 19 January, 2021 
[slides] [video]

[7] "Linearly Converging Error Compensated SGD"
Federated Learning One-World Seminar and Russian Optimization Seminar, online, 7 October, 2020
[slides] [video]

[6] "On the convergence of SGD-like methods for convex and non-convex optimization problems"
Russian Optimization Seminar, online, 8 July, 2020 (in Russian)
[slides] [video]

[5] "A Unified Theory of SGD: Variance Reduction, Sampling, Quantization and Coordinate Descent"
SIERRA, INRIA, Paris, France, 18 October, 2019
[slides]

[4] 23rd International Symposium on Mathematical Programming
Section "New methods for stochastic optimization and variational inequalities"
Talk ”An Accelerated Directional Derivative Method for Smooth Stochastic Convex Optimization”
Bordeaux, 6 July, 2018
[slides]

[3] Workshop ”Optimization at Work”
Talk ”An Accelerated Method for Derivative-Free Smooth Stochastic Convex Optimization” 
Moscow, Russia, 14 April, 2018
[slides] [video]

[2] 60th Scientific Conference of MIPT
Section of information transmission problems, data analysis and optimization
Talk ”About accelerated Directional Search with non-Euclidean prox-structure”
Moscow, Russia, 25 November, 2017
[slides]

[1] Workshop ”Optimization at Work”
Talk ”Accelerated Directional Search with non-Euclidean prox-structure”
Moscow, Russia, 27 October, 2017
[slides]