Velibor Bojkovic
Visiting Assistant Professor of Machine Learning
Research Interests
Professor Bojkovic's research interests span various areas of Machine Learning and AI, such as: Spiking Neural Networks (where his work explores various methods for training SNNs, or their theoretical and practical properties), Mechanistic interpretability of LLMs, and State Space Models. Additonally, he is interested in various properties of ramification in the context of p-adic analytic spaces and p-adic differential equations
Email
Prior to joining MBZUAI as faculty, Professor Bojkovic worked as a researcher at MBZUAI, first as a Research Fellow working on a project related to Spiking Neural Networks and then as a Research Scientist working on a project related to applications of LLMs in the energy sector. Before that, he worked as a researcher in various areas of pure mathematics. At the University of Normandie in Caen (France) he was a member of Number Theory group where he worked on non-archimedean spaces, ramification theory of local fields and theory of Drinfeld modules. Before that at University of Padova (Italy) Professor Bojkovic was member of p-adic geometry group working on p-adic differential equations and ramification theory of finite morphisms of non-archimedean curves. In addition, Professor Bojkovic worked at the Einstein Institute of Mathematics at Haifa University (Israel) in the area of non-archimedean spaces. He has a double Ph.D. degree in theoretical mathematics from the University of Padova and the University of Bordeaux.
- Research Scientist and Research Fellow at MBZUAI
- Postdoctoral position at University of Caen Normandie in France
- Postdoctoral position at University of Padova in Italy
- Postdoctoral position at Einstein Institute of Mathematics at HAIFA in Israel
- Ph.D. in theoretical mathematics at University of Padova/University of Bordeaux.
- M.Sc. in mathematics at University of Bordeaux/University of Padova
- B.Sc. in theoretical mathematics at Faculty of Mathematics and Natural Sciences in Montenegro
- ALGANT Doc European Commission schoolarship for PhD studies
- ALGANT Master European Commission scholarship for MSc studies
- V. Bojkovic, X. Wu, B. Gu, "Temporal Misalignment in ANN-SNN Conversion and Its Mitigation via Probabilistic Spiking Neurons", ICML 2025
- X. Wu, V. Bojkovic et al. "FTBC: Forward Temporal Bias Correction for Optimizing ANN-SNN Conversion", ECCV, 2024
- V. Bojkovic, S. Anumasa et al. "Data Driven Threshold and Potential Initialization for Spiking Neural Networks, AISTATS, 2024
- B. Mukhoty, V. Bojkovic et al. "Direct Training of SNNs Using Local Zeroth Order Method", NEURIPS, 2023
- V. Bojkovic, J. Poineau, "Pushforwards of p-adic differential equations", American Journal of Mathematics, 2020
- F. Baldassarri, V. Bojkovic, Metric Uniformization of morphisms of Berkovich curves via p-adic differential equations (joint with F. Baldassarri), Israel Journal of Mathematics, 2021
- V. Bojkovic, J. Poineau, "On the number of connected components of the ramification locus of a morphism of Berkovich curves", Mathematische Annalen, 2018
- V. Bojkovic, "Riemann-Hurwitz formula for finite morphisms of p-adic curves", Mathematische Zeitschrift, 2018