Reproduce “Learning to Select Sparse Linear Solvers with Convolutional Neural Networks” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Artemis Pados, Alan Edelman, Emmanuel Lujan, Daniel Pickard, Felipe Tome, Christopher Rackauckas
The authors train a convolutional neural network to recommend among UMFPACK, KLU, and MKL Pardiso using 64-by-64 images of sparse matrix nonzero patterns and benchmark timing ratios. On held-out SuiteSparse matrices, it identifies the fastest solver in 94% of cases, versus 84% for always choosing KLU. Application matrices yield 99% accuracy, though none are actual UMFPACK-fastest cases.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.