The basic research project titled “SLIMNETS – Machine Learning Networks with Sparse-Structured Local Interactions” has been approved for funding by the Hellenic Foundation for Research and Innovation (HFRI) with Professor Dionissios Hristopulos of the School of Electrical and Computer Engineering (ECE) as Principal Investigator and Scientific Lead.
Artificial Intelligence (AI) is rapidly transforming how we interpret the world. Applying AI to environmental, ecological, meteorological processes, and the assessment of natural resources requires novel machine learning methods for geospatial data. These methods must incorporate spatiotemporal (ST) correlations, efficiently scale computational resources with data size, offer accurate predictions and uncertainty quantification, operate beyond Euclidean geometries, and integrate information from multiple sources.
SLIMNETS will explore new models for spatiotemporal data using concepts from statistical field theories. It will leverage Boltzmann-Gibbs processes to develop sparse local interaction machines and use “effective medium” ideas to generate fast active regression algorithms for Gaussian processes.
SLIMNETS will design multi-output kernels based on vector Boltzmann-Gibbs processes to overcome current limitations, develop multivariate ST models with sparse equilibrium operators, and regression networks with sparse local interactions. These advancements will provide a scalable framework for spatiotemporal interpolation, prediction, and simulation, outperforming standard Gaussian processes.
SLIMNETS also aims to build computationally efficient algorithms for ST data, enhance the flexibility of Gaussian processes by creating physically-grounded multivariate kernels, and create regression models based on broader definitions of spatiotemporal distances. These innovations will support AI applications in environmental protection, early warning systems, automated pollution mapping, high-resolution soil property mapping, and filling gaps in remote sensing products.
The project is implemented under HFRI’s initiative: “3rd Call for HFRI Research Projects to Support Faculty Members and Researchers,” category II: “Mathematics and Information Sciences (Scientific Area 5),” project number HFRI 25024. Professor Athanasios Liavas of the School of ECE at the Technical University of Crete. International partners of SLIMNETS include Pete Atkinson, Executive Dean of the Faculty of Science and Technology at Lancaster University, Gerard Heuvelink, Professor at Wageningen University and Senior Researcher at ISRIC-World Soil Information, Sandra de Iaco, Professor in the Department of Economic Science at the University of Salento, and Anastasia Baxevani, Associate Professor in the Department of Mathematics and Statistics at the University of Cyprus.
This is the fifth project funded by HFRI with a Principal Investigator from the ECE School. The previous four were approved and launched in 2020, 2022, 2023, and 2025 with Principal Investigators Professor Angelos Bletsas, Associate Professor Nikolaos Bekiaris-Lymberis, Professor Georgios Chalkiadakis, and Assistant Professor Georgios Peppas, respectively.
