Sep 2026· Proceedings of the National Academy of Sciences of the United States of America· Vol 123· 1 citation· 56 references
Medicine
TL;DR
By leveraging the hidden low-rank structure of large SDPs with sparse problem data, it is demonstrated that the parallel processing power of modern graphics processing units can substantially accelerate the solution process and enable much larger problems to be solved.
Abstract
Significance Many scientific and engineering fields, including optimization, machine learning, and quantum computing, use semidefinite programming (SDP) to tackle complex problems. Unfortunately, as these problems grow, they rapidly exceed the capabilities of conventional SDP solvers, creating a critical bottleneck. By leveraging the hidden low-rank structure of large SDPs with sparse problem data, we demonstrate that the parallel processing power of modern graphics processing units can substantially accelerate the solution process and enable much larger problems to be solved. This breakthrough removes a long-standing computational barrier, enabling previously impossible experiments and optimizations across diverse areas of science and engineering.
GPU-accelerated first-order methods are increasingly effective for solving large-scale linear programs (LPs), yet LP preprocessing remains predominantly CPU-based and can become costly. A key challenge is that dependencies among presolve reductions and frequent modifications to sparse LP data make efficient GPU paralle...
Kai-Huang Chen, Jia-Wei Lu, De-Feng Sun et al.· 0 citations
PDNQP is presented, a factorization-free first-order solver for finding stationary points of large-scale nonconvex quadratic programs and its key ingredients are a residual-form reformulation that avoids penalty-dependent normal matrices and preserves a fixed sparse constraint operator across outer iterations, and an a...
This paper introduces GEM-KMeans, a spectrally normalized yet mathematically equivalent NLR formulation that fuses the gradient update, nonnegative projection, and sufficient statistics for normalization and iterate movement into a matrix-multiplication epilogue.
Peng Xu, Nihar Koganti, Volodymyr Kindratenko et al.· 0 citations
This paper introduces rlaopt, a PyTorch-based package for large-scale optimization and scientific computing using randomized numerical linear algebra (RandNLA). Despite substantial progress in RandNLA-based algorithms, few implementations combine GPU acceleration with a simple interface for specifying optimization prob...
Pratik Rathore, Zachary Frangella, P. Nobel et al.· 0 citations
A novel Jacobi-type tensor algorithm for tensor decomposition is investigated and a CUDA-based algorithm that supports tensors of arbitrary order on a single GPU is developed that outperforms the original MATLAB reference.
Markus Hellgren, Erna Begović Kovač, H. O. Karlsson et al.· 0 citations
Simple design principles that avoid severe load imbalance and expose parallelism in reductions that appear inherently sequential are described and implemented in cuPSLP, a GPU-accelerated version of PSLP, a CPU-based presolver that is itself several times faster than a state-of-the-art commercial presolver.
Daniel Cederberg, Stephen P. Boyd· 1 citation
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
