Prob-GParareal: A Probabilistic Numerical Parallel-in-Time Solver for Differential Equations
Researchers have developed Prob-GParareal, a novel probabilistic extension to the GParareal algorithm for solving differential equations. This new method incorporates Gaussian processes to model the correction function, allowing for the quantification and propagation of uncertainty across time steps. Prob-GParareal can also handle probabilistic initial conditions and integrates with existing numerical solvers. The paper demonstrates the algorithm's accuracy and robustness on various ODE systems and introduces a variant, Prob-nnGParareal, which uses nearest-neighbor Gaussian processes for improved performance on PDE examples. AI