I am a tenure-track faculty member at INSAIT, where I work on machine learning.
Dr. Nikola Konstantinov
Contact: [email protected]
Google Scholar: link
My research interests lie in the area of trustworthy machine learning and I am especially interested in providing mathematical guarantees for machine learning models. Sample of specific topics I am currently working on are:
- Making collaborative learning methods, such as federated learning, more robust and reliable;
- Studying topics at the intersection of machine learning and game theory, in particular incentives for data sharing;
- Understanding the impact of ML on society, in particular by analyzing the fairness and long-term impact of ML models.
Supervision
At INSAIT, I have the pleasure of working with the following students: Nikita Tsoy (joint with Martin Jaggi), Ivan Kirev (joint with Andreas Krause), Kristian Minchev, Kostadin Garov (joint with Martin Vechev).
Education and experience
- Postdoctoral fellow at the ETH AI Center, working under the supervision of Prof. Martin Vechev and Prof. Fanny Yang (2022-2023)
- PhD student at IST Austria, working in the group of Prof. Christoph Lampert (2017-2022). I was also part of the ELLIS PhD Program.
- Masters’ degree in Mathematics and Statistics at the University of Oxford (2013-2017)
- I attended the Sofia High School of Mathematics (2005-2013)
Awards and recognitions
- Member of the ELLIS Society since April 2023
- ETH AI Center postdoctoral fellowship in 2022
- Nomination for best PhD thesis award at IST Austria in 2022
- Member of the ELLIS PhD Program during 2021-2022
- Davies Prize awarded by Jesus College, Oxford, for the most outstanding performance in a Finals Honours School, 2017
- Department of Statistics Prize for FHS Mathematics and Statistics Part A, University of Oxford, 2015
- Bronze Medal at the Junior Balkan Mathematical Olympiad (JBMO), 2008
While at Oxford, I was also the President of the Oxford University Bulgarian Society 2015-2016.
Publications
At INSAIT
- Nikita Tsoy and Nikola Konstantinov
Simplicity Bias of Two-Layer Networks beyond Linearly Separable Data
To appear in: International Conference on Machine Learning (ICML), 2024. - Nikita Tsoy, Anna Mihalkova, Teodora Todorova and Nikola Konstantinov
Provable Mutual Benefits from Federated Learning in Privacy-Sensitive Domains
In: International Conference on Artificial Intelligence and Statistics (AISTATS), 2024 - Nikita Tsoy and Nikola Konstantinov
Strategic Data Sharing between Competitors.
In: Conference on Neural Information Processing Systems (NeurIPS), 2023. - Florian E. Dorner, Nikola Konstantinov, Georgi Pashaliev, Martin Vechev
Incentivizing Honesty among Competitors in Collaborative Learning and Optimization.
In: Conference on Neural Information Processing Systems (NeurIPS), 2023.
Before INSAIT
- Florian E. Dorner, Momchil Peychev, Nikola Konstantinov, Naman Goel, Elliott Ash, Martin Vechev
Human-Guided Fair Classification for Natural Language Processing
In: International Conference on Learning Representations (ICLR), Spotlight , 2023
Short version presented in: TSRML@NeurIPS , 2022 - Dimitar I. Dimitrov, Mislav Balunović, Nikola Konstantinov, Martin Vechev
Data Leakage in Federated Averaging
In: Transactions of Machine Learning Research (TMLR) , 2022 - Eugenia Iofinova*, Nikola Konstantinov*, Christoph H. Lampert
FLEA: Provably Fair Multisource Learning from Unreliable Training Data
In: Transactions of Machine Learning Research (TMLR) , 2022
* Denotes equal contribution - Nikola Konstantinov, Christoph H. Lampert
Fairness-Aware PAC Learning from Corrupted Data
In: Journal of Machine Learning Research (JMLR) , 2022 - Nikola Konstantinov, Christoph H. Lampert
On the Impossibility of Fairness-Aware Learning from Corrupted Data
Contributed talk + in proceedings of AFCR@NeurIPS , 2021 - Nikola Konstantinov, Elias Frantar, Dan Alistarh, Christoph H. Lampert
On the Sample Complexity of Adversarial Multi-Source PAC Learning
In: International Conference on Machine Learning (ICML), 2020 - Nikola Konstantinov, Christoph H. Lampert
Robust Learning from Untrusted Sources
In: International Conference on Machine Learning (ICML), 2019; Long Talk - Dan Alistarh, Torsten Hoefler, Mikael Johansson, Nikola Konstantinov*, Sarit Khirirat, Cedric Renggli
The Convergence of Sparsified Gradient Methods
In: Conference on Neural Information Processing Systems (NeurIPS) , 2018
* Authors’ order is alphabetical. - Dan Alistarh, Chris De Sa, Nikola Konstantinov*
The Convergence of Stochastic Gradient Descent in Asynchronous Shared Memory
In: ACM Symposium of Principles of Distributed Computing (PODC), 2018
* Authors’ order is alphabetical.
