Stochnum Ph.D. student wins 1st place - 8th KAUST-NVIDIA AI Competition

About

In October 2022, the KAUST Supercomputing Core Laboratory (KSL) in collaboration with NVIDIA promoted the 8th annual workshop on "Accelerating Scientific Applications using GPUs". 

During this event, the Spatio-temporal Statistics and Data Science (STSDS) group hosted the AI Competition on Geospatial Datasets, where the attendants competed in three categories for developing Artificial Intelligence models for analytics of big data produced by a scientific application.

Ph.D. student Yang Liu, a member of the Stochastic Numerics Research Group, attended the AI Competition. He applied the gradient boosting decision trees (GBDT) implemented in the package lightgbm to univariate (Dataset A), bivariate (Dataset B) and spatial-temporal (Dataset C) geospatial data predicting.

His prediction results achieved the smallest root-mean-squared-error and won first place in two (Dataset A and B) out of three parallel competitions.

More details about the datasets:

A - includes two (2) datasets that have been generated using univariate spatial models in 2D with 1M (106) geospatial data points. 

B - includes four (4) datasets that have been generated using bivariate  spatial models in 2D with 500K geospatial data points. 

C -  includes nine (9) datasets that have been generated using spacetime models in 2D×time with 10K geospatial data points and 100-time points. 

Stochnum team congratulates Yang Liu for his great performance and celebrates the victory!