Talks and Poster Presentations
2026
- Swiss Optimization SymposiumAugust 23–27, 2026Monte Verità · Ascona, SwitzerlandPoster: Rescaled Asynchronous SGD: Optimal Distributed Optimization under Data and System Heterogeneity [poster]
- Protocol Learning: Decentralized Collaborative Learning at Scale (ICLR 2026 Workshop) April 26, 2026Lagune Barra Hotel · Rio de Janeiro, Brazil
- Riocentro Convention and Event Center · Rio de Janeiro, BrazilPoster: Ringleader ASGD: The First Asynchronous SGD with Optimal Time Complexity under Data Heterogeneity [poster]
- KAUST Rising Stars in AI Symposium 2026 February 9, 2026KAUST, Saudi ArabiaPosters:
2025
- AMCS/STAT graduate seminar November 13, 2025KAUST, Saudi ArabiaTalk: First Provably Optimal Asynchronous SGD for Homogeneous and Heterogeneous Data [slides]
- The KAUST 2025 Workshop on Statistics November 4, 2025KAUST, Saudi Arabia
- Mathematics and Applications Colloquium October 28, 2025KAUST, Saudi ArabiaTalk: Ringleader ASGD: The First Asynchronous SGD with Optimal Time Complexity under Data Heterogeneity [slides]
- 35º Colóquio Brasileiro de Matemática July 27–August 1, 2025Instituto Nacional de Matemática Pura e Aplicada (IMPA) · Rio de Janeiro, Brazil
- 41st Conference on Uncertainty in Artificial Intelligence July 21–25, 2025Rio Othon Palace · Rio de Janeiro, BrazilPoster: MindFlayer SGD: Efficient Parallel SGD in the Presence of Heterogeneous and Random Worker Compute Times [poster]
- Forty-Second International Conference on Machine Learning July 13–19, 2025Vancouver Convention Center · Vancouver, Canada
- Stochastic Numerics and Statistical Learning May 18, 2025KAUST, Saudi ArabiaPoster: ATA: Adaptive Task Allocation for Efficient Resource Management in Distributed Machine Learning [poster]
- The Thirteenth International Conference on Learning Representations April 24–28, 2025Singapore EXPO · SingaporePoster: LoCoDL: Communication-Efficient Distributed Learning with Local Training and Compression (ICLR 2025 Spotlight, top 5.1%) [poster]
- Federated Learning One World Seminar (FLOW) April 16, 2025FLOW Talk #126 · Online
- KAUST Rising Stars in AI Symposium 2025 April 7–10, 2025KAUST, Saudi ArabiaPoster: ATA: Adaptive Task Allocation for Efficient Resource Management in Distributed Machine Learning [poster]
- Machine Learning Reading Group March 28, 2025YSU Krisp-AI Lab · Yerevan, ArmeniaTalk: Ringmaster ASGD: The First Asynchronous SGD with Optimal Time Complexity [slides]
- Flower AI Summit 2025 March 26, 2025King's House · London, EnglandTalk: Ringmaster ASGD: The First Asynchronous SGD with Optimal Time Complexity [slides]
- Academic Report of the School of Mathematical Sciences March 17, 2025Beihang University (BUAA) · Beijing, ChinaInvited by Jiaxin Xie to give a talk on Ringmaster ASGD: The First Asynchronous SGD with Optimal Time Complexity [slides]
- Optimization Seminar March 13, 2025Beijing Institute of Mathematical Sciences and Applications (BIMSA) · Beijing, ChinaInvited by Yi-Shuai Niu to give a talk on Ringmaster ASGD: The First Asynchronous SGD with Optimal Time Complexity [slides]
- Seminar March 12, 2025Peking University · Beijing, ChinaInvited by Kun Yuan to give a talk on Ringmaster ASGD: The First Asynchronous SGD with Optimal Time Complexity [slides]
- AMCS/STAT graduate seminar February 27, 2025KAUSTTalk: Ringmaster ASGD: The First Asynchronous SGD with Optimal Time Complexity [slides]
2024
- Workshop on Optimization for Machine Learning (NeurIPS 2024) December 15, 2024Vancouver Convention Center · Vancouver, CanadaPresented
- MLR Weekly SeminarNovember 21, 2024Machine Learning Research at Apple · OnlineInvited by Samy Bengio to give a talk on MindFlayer: Efficient Asynchronous Parallel SGD in the Presence of Heterogeneous and Random Worker Compute Times [slides]
- International Conference on Algebra, Logic, and their Applications October 18, 2024Yerevan State University · OnlineTalk: MindFlayer: Efficient Asynchronous Parallel SGD in the Presence of Heterogeneous and Random Worker Compute Times
- CEMSE E-Poster Competition October 10, 2024KAUST, Saudi ArabiaPoster: GradSkip: Communication-Accelerated Local Gradient Methods with Better Computational Complexity [3rd place]
- Analysis, PDEs and Applications July 6, 2024Yerevan State University · Yerevan, ArmeniaTalk: MindFlayer: Efficient Asynchronous Parallel SGD in the Presence of Heterogeneous and Random Worker Compute Times [abstract]
- Stochastic Numerics and Statistical Learning May 27, 2024KAUST, Saudi ArabiaPoster: GradSkip: Communication-Accelerated Local Gradient Methods with Better Computational Complexity [poster]
- CS 331, Stochastic Gradient Descent MethodsMay 5, 2024KAUST, Saudi ArabiaDelivered a guest lecture on MindFlayer: Efficient Asynchronous Parallel SGD in the Presence of Heterogeneous and Random Worker Compute Times
- The Machine Learning Summer School in Okinawa 2024 March 13, 2024Okinawa Institute of Science and Technology (OIST) · Okinawa, JapanPoster: GradSkip: Communication-Accelerated Local Gradient Methods with Better Computational Complexity [poster]
- KAUST Rising Stars in AI Symposium 2024 February 21, 2024KAUST, Saudi ArabiaPoster: GradSkip: Communication-Accelerated Local Gradient Methods with Better Computational Complexity [poster]
2023
- Group SeminarNovember 16, 2023KAUST, Saudi ArabiaTalk: Differentially Private Coordinate Descent for Composite Empirical Risk Minimization
- Algorithms & Computationally Intensive Inference seminars October 6, 2023University of Warwick · Coventry, EnglandTalk: GradSkip: Communication-Accelerated Local Gradient Methods with Better Computational Complexity [slides]
- Mathematics in Armenia: Advances and Perspectives July 5, 2023Yerevan State University · Yerevan, ArmeniaTalk: GradSkip: Communication-Accelerated Local Gradient Methods with Better Computational Complexity [abstract]
- Machine Learning Reading Group Yerevan March 10, 2023Yerevan State University · Yerevan, ArmeniaTalk: GradSkip: Communication-Accelerated Local Gradient Methods with Better Computational Complexity [video (Armenian)]
2022
- Federated Learning One World Seminar (FLOW) December 7, 2022FLOW Talk #88 · OnlineTalk: GradSkip: Communication-Accelerated Local Gradient Methods with Better Computational Complexity [video]
- Machine Learning Reading Group Yerevan April 10, 2022Yerevan State University · Yerevan, ArmeniaTalk: ProxSkip: Yes! Local Gradient Steps Provably Lead to Communication Acceleration! Finally!