Reproduce “Scalable Extreme Deconvolution” 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
James A. Ritchie, Iain Murray
Ritchie and Murray make extreme deconvolution scalable by fitting noise-aware Gaussian mixtures with minibatch EM or stochastic gradient descent (SGD), rather than requiring full-dataset updates. They evaluate the methods on a two-million-row Gaia subset, with EM on CPU and minibatch methods on GPUs. At K=512, SGD achieves held-out average log-likelihood −25.57 versus −25.66 for EM; differing regularization limits comparison.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.