Reproduce “CloudLSTM: A Recurrent Neural Model for Spatiotemporal Point-cloud Stream Forecasting” 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

Chaoyun Zhang, Marco Fiore, Iain Murray, Paul Patras

CloudLSTM forecasts values at irregularly located sensors without mapping them onto a grid. Its DConv operation learns local neighborhoods from spatial coordinates, preserves input-order invariance, and updates point features within recurrent forecasting. In the 12-step air-quality evaluation, CloudLSTM with K=3 reports mean MAE 9.12 versus 10.39 for ConvLSTM; this comparison applies to the tested datasets and setting.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.