Rustem Islamov
About me
I am a third 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.
On top of that, I am a big fan of tennis and judo.
Education
PhD in Computer Science
Supervisor: Aurelien Lucchi
Master of Science in Data Science
Thesis: Unified Analysis of Asynchronous Algorithms
Thesis Supervisor: Mher Safaryan, Dan Alistarh
Bachelor of Science in Applied Mathematics and Physics
Thesis: Distributed Second Order Methods with Fast Rates and Compressed Communication
Thesis Supervisor: Peter Richtárik
Contacts
- rustem.islamov@unibas.ch
- Spiegelgasse 1, 4051 Basel, Switzerland
- GitHub
- Google Scholar
- CV
Interests
- Optimization
- Optimization Theory
- Computation/Memory-efficient and Scalable Algorithms
- Convex/Non-convex Stochastic Optimization
- Machine Learning
- Loss Landscape of Neural Networks
- Federated learning
- Communication-efficient and Scalable Algorithms
Selected Publications
My contributions have appeared in leading machine learning venues including NeurIPS, ICML, ICLR, and AISTATS. Here is the list of papers I am most proud of:
[21]
On the Role of Batch Size in Stochastic Conditional Gradient Methods, (ICML 2026).
[20]
Non-Euclidean Gradient Descent Operates at the Edge of Stability, (Oral, ICML 2026)
[14]
Unbiased and Sign Compression in Distributed Learning: Comparing Noise Resilience via SDEs, (Oral, AISTATS 2025).
Recent news
- June 6, 2026I attended SIOPT 2026. I organized a session on recent advances in federated optimization.
- June 1, 2026One paper got accepted to UAI 2026 in Amsterdam.
- May 8, 2026I attended ELLITE Focus Period in Lund as a visiting scholar. I spent two wonderful weeks in Lund discussing optimization for machine learning.
- May 1, 2026Four papers accepted to ICML 2026, including an Oral.
- Jan 20, 2026Paper on analysis of DP algorithms with SDEs was accepted to ICLR 2026.