Reproduce “Rotation-Invariant Random Features Provide a Strong Baseline for Machine Learning on 3D Point Clouds” 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
Owen Melia, Eric Jonas, Rebecca Willett
The authors present a rotation-invariant model for 3D point clouds, using randomized features and a linear readout instead of a neural network. On QM7, it reports 0.0704 eV mean absolute error and 0.0189 seconds prediction latency, versus FCHL19’s 0.0705 eV and 0.6635 seconds. On QM9, random features and Cormorant report 0.022 eV, but the method is not state-of-the-art and its approximate solve takes 76.5 hours.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.