GP Inference

Gaussian process (GP) inference aims to efficiently perform probabilistic predictions using GP models, which are powerful but computationally expensive for large datasets. Current research focuses on developing scalable algorithms, such as those employing coresets, state-space representations, and sparse approximations like the product of experts, to reduce the computational complexity from cubic to linear time. These advancements enable the application of GPs to large-scale problems across diverse fields, including robotics, solar power forecasting, and pedestrian trajectory prediction, where accurate uncertainty quantification is crucial.