A Neural Hierarchical-Matrix Preconditioner for Real-Time GPU Solves
Abstract
Interactive simulation solves Ax=b for a sparse SPD A that changes every frame, inside an 8-16 ms budget. At a few thousand unknowns, the setup of algebraic multigrid alone exceeds that budget, while Jacobi and other local preconditioners have no setup but cannot move error across the domain. We learn a preconditioner for this gap: a graph-and-attention network predicts an SPD approximate inverse in H^2-matrix format. On a spatially ordered 3D mesh, blocks of the true inverse lose rank as the clusters they couple move apart; the nested bases of the format follow that decay, so inference and apply are dominated by leaf-block work linear in N, where a dense inverse costs N^2. Our main finding concerns training. Probe losses reach M only through a product with A, so their gradient vanishes on the near-null modes that set the conjugate-gradient iteration count. A truncated Kaporin condition number has no such factor; changing only the objective cuts iterations on a held-out frame from 116 to 33. On a ladder of stiff tetrahedral diffusion problems ours alone fits an 8.3 ms (120 fps) frame from N=572 to 3,647.
Versions
Related
- Singular perturbations and hierarchical learning in two-layer neural networks
- New hierarchical matrices for fast kernel matrix-vector products with Applications to Gaussian processes and kernel ridge regression
- HiLNO: A Hierarchical Latent Neural Operator with Multi-Scale Supervision for PDEs on General Geometries
- A Two-Stage Optimization Framework for Real-World Large-Scale University Course Timetabling Based on Orthogonalized Hierarchical Modeling and Conflict-Driven ALNS Repair
- Factorisation and matrix amplification of real rank zero inclusions
BibTeX
@misc{osborne2026neural,
title = {{A Neural Hierarchical-Matrix Preconditioner for Real-Time GPU Solves}},
author = {Carl Osborne and Minghao Guo and Crystal Owens and Wojciech Matusik},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2605.13343},
eprint = {2605.13343},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2605.13343},
}
RIS
TY - UNPB
TI - A Neural Hierarchical-Matrix Preconditioner for Real-Time GPU Solves
AU - Carl Osborne
AU - Minghao Guo
AU - Crystal Owens
AU - Wojciech Matusik
PY - 2026
DA - 2026/09/24
PB - arXiv
DO - 10.48550/arxiv.2605.13343
UR - https://arxiv.org/abs/2605.13343
AB - Interactive simulation solves Ax=b for a sparse SPD A that changes every frame, inside an 8-16 ms budget. At a few thousand unknowns, the setup of algebraic multigrid alone exceeds that budget, while Jacobi and other local preconditioners have no setup but cannot move error across the domain. We learn a preconditioner for this gap: a graph-and-attention network predicts an SPD approximate inverse in H^2-matrix format. On a spatially ordered 3D mesh, blocks of the true inverse lose rank as the clusters they couple move apart; the nested bases of the format follow that decay, so inference and apply are dominated by leaf-block work linear in N, where a dense inverse costs N^2. Our main finding concerns training. Probe losses reach M only through a product with A, so their gradient vanishes on the near-null modes that set the conjugate-gradient iteration count. A truncated Kaporin condition number has no such factor; changing only the objective cuts iterations on a held-out frame from 116 to 33. On a ladder of stiff tetrahedral diffusion problems ours alone fits an 8.3 ms (120 fps) frame from N=572 to 3,647.
ER -