Abstract: Large-scale simulations and parallel computing techniques are becoming essential in Gaussian process calculations to lessen the complexity of geostatistics applications. The log-likelihood function is used in such applications to evaluate the model associated with a given set of measurements in existing n geographic locations. The evaluation of such a function requires O(n^2) memory and O(n^3) computation, which is infeasible for large datasets with existing software tools.
We present ExaGeoStatR, a package for large-scale geostatistics in R that computes the log-likelihood function on shared and distributed-memory, possibly equipped with GPU, using advanced linear algebra techniques. The package provides a high-level abstraction of the underlying architecture while enhancing the R developers' productivity. We demonstrate ExaGeoStatR package by illustrating its implementation details, analyzing its performance on various parallel architectures, and assessing its accuracy using synthetic datasets and a sea surface temperature dataset. The performance evaluation involves spatial datasets with up to 250K observations.
Best Poster Finalist (BP): no
Poster summary: PDF
Back to Poster Archive Listing