I am a first year PhD student at University of Basel working under the supervision of Aurelien Lucchi. Prior to that, I obtained my Master diploma from Institut Polytechnique de Paris in Data Science and Bachelor diploma from Moscow Institute of Physics and Technology in Applied Mathematics. I am interested in Optimization and its applications to Machine Learning.

Education:

• PhD Candidate in Computer Science
  • Supervisor: Aurelien Lucchi
  • Thesis: Optimization for Machine Learning
  • Oct. 2023 - Present
• Master of Science in Data Science
  • Supervisor: Mher Safaryan, Dan Alistarh 
  • Thesis: AsGrad: A Sharp Unified Analysis of Asynchronous-SGD Algorithms
  • Sept. 2021 - Sept. 2023
• Bachelor of Science in Applied Mathematics
  • Supervisor: Peter Richtárik
  • Thesis: Distributed Second Order Methods with Fast Rates and Compressed Communication
  • Sept. 2017 - July 2021

Publications:

[9] Yuan Gao*, Rustem Islamov*, Sebastian Stich, EControl: Fast Distributed Optimization with Compression and Error Control, arXiv preprint arXiv: 2311.05645, 2023.

[8] Rustem Islamov, Mher Safaryan, Dan Alistarh. AsGrad: A Sharp Unified Analysis of Asynchronous-SGD Algorithms, arXiv preprint arXiv: 2310.20452, 2023.

[7] Konstantin Mishchenko, Rustem Islamov, Eduard Gorbunov, Samuel Horváth. Partially Personalized Federated Learning: Breaking the Curse of Data Heterogeneity, arXiv preprint arXiv: 2305.18285, 2023.

[6] Sarit Khirirat, Eduard Gorbunov, Samuel Horváth, Rustem Islamov, Fakhri Karray, Peter Richtarik. Clip21: Error Feedback for Gradient Clipping, arXiv preprint arXiv: 2305.18929, 2023.

[5] Maksym Makarenko, Elnur Gasanov, Rustem Islamov, Abdurahmon Sadiev, Peter Richtárik. Adaptive Compression for Communication-Efficient Distributed Training, Transactions on Machine Learning Research, 2022.

[4] Rustem Islamov, Xun Qian, Slavomír Hanzely, Mher Safaryan, Peter Richtárik. Distributed Newton-Type Methods with Communication Compression and Bernoulli Aggregation, Transactions on Machine Learning Research, 2022.

[3] Xun Qian, Rustem Islamov, Mher Safaryan, Peter Richtárik. Basis Matters: Better Communication-Efficient Second Order Methods for Federated Learning, in Proc. of the 25th International Conference on Artificial Intelligence and Statistics, 2022.

[2] Mher Safaryan, Rustem Islamov, Xun Qian, Peter Richtarik. FedNL: Making Newton-Type Methods Applicable to Federated Learning, In Proc. of 39th International Conference on Machine Learning, 2022.

[1] Rustem Islamov, Xun Qian, Peter Richtárik. Distributed Second Order Methods with Fast Rates and Compressed Communication, In Proc. of 38th International Conference on Machine Learning, 2021. 

* denotes equal contributions.

Last updated: 17^th of November 2023