A presentation titled “On space-time local interaction machines” was delivered at the 19th International Joint Conference CFE-CMStatistics 2025, held at Birkbeck, University of London. The talk was presented by Prof. Dionissios Hristopulos from the Technical University of Crete (Greece).
The presentation focused on the development of space–time local interaction machines (SLIMs) for scalar and vector (multi-output) spatiotemporal processes. SLIMs are inspired by statistical physics models and leverage sparse precision matrices, enabling significantly improved computational efficiency in regression tasks compared to Gaussian process frameworks. The talk discussed the construction of positive-definite precision matrices capable of incorporating general distance measures, presented parameter estimation and regression formulations based on these matrices, and demonstrated the approach using example datasets.