Guram Mikaberidze

Selected publications

In this project, we studied the dynamics of flows on networks under various symmetries and constraints. We developed a hierarchy of field theoretic tools on networks, capturing emergent behaviors of packets flowing through networks with routing strategies spanning from random walk to shortest path. Notably, the novel mesoscopic field theoretic approach admits a closed form solution by aggregating irrelevant aspects of network structure, and provides analytical insights into this complex nonlinear dynamics.
Mikaberidze, G., Artime, O., Díaz-Guilera, A., & D'Souza, R. M. (2025). Multiscale field theory for network flows. Physical Review X, 15(2), 021044. https://doi.org/10.1103/PhysRevX.15.021044

This project aims to preserve the Kartvelian language family—starting with Georgian—by developing NLP tools for this low-resource, morphologically rich language. Our latest work delivers the first in-depth evaluation of tokenization methods for Georgian, comparing standard approaches and analyzing their performance across key NLP tasks to uncover practical trends and guidance on what to use when.
Mikaberidze, B., Saghinadze, T., Mikaberidze, G., Kalandadze, R., Pkhakadze, K., van Genabith, J., Ostermann, S., van der Plas, L. and Müller, P., 2024, October. A Comparison of Different Tokenization Methods for the Georgian Language. In Proceedings of the 7th International Conference on Natural Language and Speech Processing (ICNLSP 2024) (pp. 199-208).

In this project we explored diverse mobile agents moving through and interacting in a complex network of different environments, similar to how people interact in various physical and virtual spaces. The internal workings of these agents and their interactions are flexible to cater to various applications. The study results in effective equations for understanding agent behavior, useful for examining emergence of consensus, creating strategies to disrupt cohesion, and optimizing network structures. Applications of these findings include studying synchronized brain activity in group settings (as observed in recent MRI experiments) or understanding how consensus forms in opinion dynamics. The animation illustraties Kuramoto agents moving through a network of locations that facilitate various coupling strengths.
Mikaberidze, G., Chowdhury, S.N., Hastings, A. and D'Souza, R.M., 2023. Consensus Formation Among Mobile Agents in Networks of Heterogeneous Interaction Venues. Chaos, Solitons & Fractals, 178, 114298.
Project code is available on GitLab, preprint on arXiv.

This recent study focuses on self-organized criticality (SOC) - a phenomenon where systems naturally converge to rare, scale invariant states. While many real-world scenarios display SOC, characterized by frequent and large fluctuations, many of them display even larger numbers of devastating events than scale invariance would predict. These eveents are known as dragon kings (DK). We explored how DK events emerge in SOC models and discovered that the interplay between driving impulse and energy dissipation plays a crucial role. By tweaking this balance, we found ways to minimize these disruptions back to the scale-invariant expectations and we analytically found the boundary between the two operating regimes. The figure shows the classification of DK events resulting from our study.
Mikaberidze, G., Plaud, A and D'Souza, R.M., 2023. Dragon kings in self-organized criticality systems. Physical Review Research, 5, L042013 (Editors' suggestion).
Project code is available on GitLab, preprint is on arXiv.

In this work, we studied the emergent properties that arise when two seminal models in dynamical systems, the Bak-Tang-Weisenfeld (BTW) model of sandpile cascades and the Kuramoto model of synchronization in oscillator networks, are coupled together. The BTW model captures cascades of load shedding and has been used to model brain networks and electric power grids, but it does not capture the key fact that the nodes in both of these networks are oscillators. We study the interplay between cascading loads and the synchronization of oscillators and show that it leads to emergent new behaviors and long time-scale organization. It is common to liken the large events in the BTW model to “black swans” which are massive events generated by the same underlying mechanism as the smaller events. In contrast, the chain of cascades in our model can be likened to a “dragon king” which is a massive event that occurs significantly more often than what extrapolation from smaller events would suggest. The presented animation illustrates the coupled BTW-Kuramoto model in action.
Mikaberidze, G. and D'Souza, R.M., 2022. Sandpile cascades on oscillator networks: The BTW model meets Kuramoto. Chaos: An Interdisciplinary Journal of Nonlinear Science, 32(5), p.053121 (Editor's pick).
Project code is available on GitLab, preprint is on arXiv.

Conformal Field Theory (CFT) is a type of Quantum Field Theory that is invariant under conformal transformations, that is, generalized scale invariance. CFT is a powerful tool with applications in string theory, quantum gravity, and critical phenomena in statistical physics. The notion of universality classes in statistical physics is closely related to the strict constraints on self-consistent CFTs, and Conformal Bootstrap utilizes this self-consistency to tighten the limits around possible CFTs. In this project, we formulated the conformal bootstrap equations in momentum space and demonstrated that they are convergent.
Gillioz, M., Lu, X., Luty, M.A. and Mikaberidze, G., 2020. Convergent momentum-space OPE and bootstrap equations in conformal field theory. Journal of High Energy Physics, 2020(3), pp.1-22.
Preprint is available on arXiv.

Plasma, the fourth fundamental state of matter characterized by a large fraction of charged particles, can become degenerate under high pressures where quantum effects, such as Pauli exclusion, become important. Highly relativistic plasmas are observed in the cores of white dwarfs, in magnetosphere of pulsars, in the MeV epoch of the early Universe, and probably in Active Galactic Nuclei. We found standing electro-magnetic solitons by numerically analyzing relativistic hydrodynamics and Maxwell equations.
Mikaberidze, G. and Berezhiani, V.I., 2015. Standing electromagnetic solitons in degenerate relativistic plasmas. Physics Letters A, 379(42), pp.2730-2734..
Preprint is available on arXiv.