4-year PhD position in Theoretical condensed matter / Computational physics

Topic: Application of spin models in spatial data prediction

The research will be devoted to the use of appropriately defined spin models for the prediction of missing values in predominantly massive space-time data, e.g. from remote sensing of the Earth. Traditional prediction methods are not suitable for such data, mainly due to high computational complexity as well as other limitations [1]. Recent research has shown that prediction methods, based for example on the application of a simply modified planar rotator spin model [2,3] and its generalized version [4], can be computationally much more efficient. The potential of spin models lies in the ability to model various types of time-space correlations using short-range inter-spin interactions along with global external effects. The proposed research aims to develop strategies for the development of efficient prediction methods that would be flexible and suitable for automatic processing using massively parallel algorithms implemented on graphics processors (GPUs).

Literature:

1.N. Cressie and C.K. Wikle, Statistics for spatio-temporal data. John Wiley & Sons, 2015.

2.M. Žukovič and D.T. Hristopulos, Gibbs Markov random fields with continuous values based on the modified planar rotator model, Phys. Rev. E 98 062135 (2018).

3.M. Žukovič, M. Borovský, M. Lach and D.T. Hristopulos, GPU-Accelerated Simulation of Massive Spatial Data Based on the Modified Planar Rotator Model, Mathematical Geosciences 52 123 (2020).

4.M. Žukovič and D.T. Hristopulos, Spatial data modeling by means of Gibbs Markov random fields based on a generalized planar rotator model, Physica A 612 128509 (2023).

Contact: prof. Milan Zukovic, PhD. ([email protected])