Reproduce “GPUMAP: GPU-Accelerated Progressive Uniform Manifold Approximation and Projection” 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
Myeongwon Jung, Jaemin Jo
GPUMAP splits UMAP into GPU-batched INSERT, UPDATE, and EMBED operations scheduled under time budgets, allowing projections to appear during computation. On 5-million-point HIGGS, it produced intermediate projections at 8.42, 19.86, and 27.03 seconds, before standard UMAP’s complete result at 50.4 seconds. This shows earlier access, not equivalent quality or faster completion; progressive execution can increase total runtime.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.