AI & ML
Robust BUG dynamical low-rank integrators based on the Euler-Maruyama method for stochastic differential equations. Part I: a closed-moment approach
arXiv cs.NA0 views
arXiv:2611.25152v1 Announce Type: new Abstract: In this work, we introduce Euler-Maruyama-based Basis-Update & Galerkin (EM-BUG) dynamical low-rank integrators for stochastic differential equations (SDEs). These integrators extend the standard BUG integrator to the Euler-Maruyama scheme by incorporating diffusion directions into the augmented physical basis and exploiting the second moment to update the physical basis independently of the new Brownian increments. This construction avoids the inversion of the reduced second-moment matrix and allows us to derive convergence estimates that are robust with respect to small eigenvalues. We establish preservation of positive semidefiniteness, exactness with respect to Euler-Maruyama at full rank, and second-moment stability. Strong and weak convergence estimates show that the time-discretization accuracy of the underlying Euler-Maruyama scheme is preserved up to the low-rank approximation error. We also derive Monte Carlo error estimates recovering the standard sampling rate up to the time-discretization and low-rank approximation errors. Numerical experiments on linear SDEs with additive and multiplicative noise support these theoretical results and demonstrate accuracy comparable to the standard BUG integrator.
Source:https://arxiv.org/abs/2611.25152PreviousHEL: A Heterogeneous Extensible Layer for GPU-Resident Runtime OrchestrationNextA fast iterative solver for lopsided Sylvester matrix equations
Related posts

On modularization
0dev.to
Is there a standard for AI agent receipts? Yes. Five systems bent around it in one day.
0dev.to
Installed but never connects: five platforms, five different permission prompts
0dev.to
465 practice tests, 423 pages, and every single test without a page of its own is a full sitting
0dev.to
Sort a Month of Stripe Failures by Code Before You Change Anything
0dev.to
How we find viral moments in a 6-hour stream in 4 minutes
0dev.to